Microsoft Photo Editor - omslagfront.jpg

Transcription

DYNAMICS AND DESIGN OPTIMISATION OF
OFFSHORE WIND ENERGY CONVERSION SYSTEMS
PROEFSCHRIFT
ter verkrijging van de graad van doctor
aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus prof. ir. K.F. Wakker,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen op dinsdag 5 juni 2001 om 10.30 uur
door
Martin Johannes KÜHN
‘Diplom-Ingenieur Physikalische Ingenieurwissenschaft’
Technische Universiteit Berlijn, Duitsland
geboren te Linne nu Bissendorf, Duitsland
Dit proefschrift is goedgekeurd door de promotoren:
Prof. dr. J.B. Dragt, Technische Universiteit Delft
Prof. Dr.-Ing. R. Gasch, Technische Universiteit Berlijn
Toegevoegd promotor:
Dr. G.J.W. van Bussel, Technische Universiteit Delft
Samenstelling promotiecommissie:
Rector Magnificus, voorzitter
Prof. dr. J.B. Dragt
Prof. Dr.-Ing. R. Gasch
Dr. G.J.W. van Bussel
Prof. dr. ir. G.A.M. van Kuik
Prof. dr. ir. J.H. Vugts
Prof. dr. ir. J. Wardenier
D. Quarton, M.A.
Technische Universiteit Delft
Technische Universiteit Berlijn
Garrad Hassan & Partners, Bristol
ISBN 90-76468-07-9
DUWIND Delft University Wind Energy Research Institute, Report 2001.002
2nd Edition with minor corrections, Feb. 2003
Keywords: Wind Energy, Offshore, Dynamics, Design Optimisation, Fatigue
Published and distributed by the author in cooperation with:
DUWIND Delft University Wind Energy Research Institute
Stevinweg 1, 2628 CN Delft, The Netherlands
Phone: +31 15 278 5170
http://www.duwind.tudelft.nl
Cover design:
Main cover photo:
Reproduction:
LenShape creative products, www.lenshape.nl
Jos Beurskens, ECN
Doc Vision BV. Technische Universiteit Delft
Copyright © 2001, by M. Kühn
Von-Velen-Weg 11, D-48431 Rheine
Email: m.kuehn.rheine@t-online.de
All rights reserved. Any use or application of data, methods and/or results etc.,
occurring in this thesis will be at user's own risk. The author accepts no liability for
damage suffered from the use or application.
No part of the material protected by the copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical, including
photocopying, recording or by any information storage and retrieval system,
without permission of the author.
Stellingen
behorend bij het proefschrift
Dynamics and Design Optimisation of
Offshore Wind Energy Conversion Systems
van
Martin Kühn
1.
Voor grootschalige exploitatie van windenergie buitengaats is de ontwikkeling
van een volwassen offshore windenergietechnologie van cruciaal belang. Basis
voor
een
dergelijke
ontwikkeling
zijn
de
ervaringen
met
de
windenergietechnologie op het land, de offshore technologie en de
bedrijfsvoering van grote elektriciteitscentrales.
Dit proefschrift.
2.
Het kruien van de windturbine uit de windrichting in de golfrichting om de
vermoeiingsbelastingen van een offshore windenergie-converter te verlagen,
zoals voorgesteld door Sinclair en Clayton, zal het tegengestelde effect hebben.
Sinclair, F. M.; Clayton, B. R., 1989, Excitation and Damping Forces on Offshore Wind
Turbines. Wind Engineering 13, 6, 276 - 292.
3.
Op overtrek geregelde windturbines met een vrijwel vast toerental zijn niet
geschikt voor grootschalige toepassing buitengaats.
4.
Op locatie moeten offshore windenergie-converters uit een minimaal aantal
hoofdonderdelen worden opgebouwd. Over vijf à tien jaar zal, afhankelijk van de
configuratie, dit aantal één tot drie bedragen.
5.
Op langere termijn zullen de tweebladige offshore windenergie-converters
herleven.
6.
Kwaliteit is belangrijker dan kwantiteit bij de ontwikkeling van windparken
buitengaats.
7.
De snelle technologische ontwikkeling en groei van de internationale
windenergie-industrie belemmert op dit moment de belangstelling voor zowel
fundamenteel als toegepast onderzoek.
8.
De suggestie van Bongers dat het optimaal gebruik van technologische kennis
bepalend zou zijn voor het succes van de Nederlandse windturbinefabrikanten
is noch in Nederland noch in verschillende andere landen juist gebleken.
Stelling 12 van Bongers, P.M.M., 1996, Modeling and Identification of Flexible Wind
1
Turbines and Factorizational Approach to Robust Control. TU Delft.
9.
Directe stimulering van studenten en jonge onderzoekers is belangrijker voor de
toekomst van de Technische Universiteit Delft dan de aanstelling van
toponderzoekers.
10. Internetten op school is niet cruciaal voor een succesvolle opvoeding.
11. De toevoeging van funnies in hagelslag bevordert de omzet. Het verdient
aanbeveling deze verkoopmethode vooral bij verantwoorde voedingsmiddelen
toe te passen.
1
‘De Nederlandse windturbinefabrikanten zullen alleen concurrerend kunnen produceren indien zij
optimaal gebruik gaan maken van de technologische know-how.’
Propositions
belonging to the thesis
Dynamics and Design Optimisation of
Offshore Wind Energy Conversion Systems
by
Martin Kühn
1.
For large-scale exploitation of offshore wind energy, the development of a
mature offshore wind energy technology is of crucial importance. Such
development should be based on the experiences with onshore wind energy
technology, offshore technology and operation of large power plants.
This thesis.
2.
The proposal of Sinclair and Clayton, to reduce the fatigue loads of an offshore
wind energy converter by yawing the turbine from the wind direction into the
wave direction, will have the opposite effect.
Sinclair, F. M.; Clayton, B. R., 1989, Excitation and Damping Forces on Offshore Wind
Turbines. Wind Engineering 13, 6, 276 - 292.
3.
Stall regulated wind turbines, with almost constant rotor speed, are not suitable
for large-scale offshore application.
4.
Offshore wind energy converters have to be installed onsite with a minimum
number of pre-assembled sections. Within five to ten years this number will be
between one and three, depending on the configuration.
5.
In the long-term, two-bladed offshore wind energy converters will revive.
6.
Quality is more important than quantity for the development of offshore wind
farms.
7.
The rapid technological progress and growth of the international wind energy
industry hinders the interest for both fundamental and applied research today.
8.
Bongers stated, that the optimum application of technological know-how would
be decisive for the success of the Dutch wind turbine manufacturers. Neither in
The Netherlands nor in other countries this turned out to be true.
Proposition 12 of Bongers, P.M.M., 1996, Modeling and Identification of Flexible Wind
Turbines and Factorizational Approach to Robust Control. TU Delft.
9.
The direct stimulation of students and young researchers will be more important
for the future of the Delft University of Technology than the employment of top
researchers.
10. Internet use at school is not crucial for a successful education.
11. The addition of ‘funnies’ to chocolate sprinkles (‘hagelslag’) stimulates sales. It
is recommended to apply this sales approach to healthy food, in particular.
M. Kühn - Dynamics and Design Optimisation of Offshore Wind Energy Conversion Systems
SUMMARY
Offshore wind energy will become an important renewable energy source for several
countries in Northern Europe. Currently, rapid developments are taking place, with the
technical and general economic feasibility having been demonstrated by small-scale
prototype projects in sheltered waters during the 1990's. In 2000, the first offshore wind
energy converters in the megawatt class were installed, a technology which will also be
employed for large-scale offshore wind farms with 100 to 160 MW capacity by the year
2002. For the medium and long term, multi-megawatt turbines and offshore wind farms
in the range up to one gigawatt have been announced.
For exploitation of offshore wind energy on such scale, the development of a mature
offshore wind energy technology is of crucial importance. Such development should be
based on the experiences with onshore wind energy technology, offshore technology of
the petroleum industry and operation of large power plants.
How can analysis of the structural dynamics and a specific design approach, improve
the cost-efficiency and reliability of offshore wind energy conversion systems?
This question is the subject of the present Doctoral Thesis and is answered by four
contributions:
• Development and demonstration of a design methodology for offshore wind farms
• Synthesis of analysis methods for the structural dynamics; the complexity of which
is adjusted to the different stages of the design process
• Investigation of the particular dynamics of offshore wind energy converters and the
interrelation with the design
• Demonstration of a design optimisation by means of tailored dynamics
In the following, the main results in these fields are sketched.
After a general introduction and a summary of the state of the art, the integrated design
approach for offshore wind farms is introduced. This methodology considers all
components as wind turbine, support structure and grid integration as well as installation
and operation and maintenance aspects during the design decisions. Interactions
between subsystems are considered as completely and practically as possible. Hence,
the design solution is governed by overall criteria including levelised production costs,
dynamics of the entire system, adaptation to the actual site conditions, installation effort
and wind farm availability.
A reference approach for the analysis of bottom-mounted offshore wind energy
converters is synthesised from the expertise in both onshore wind energy technology
and offshore technology. This enables investigation of the response under simultaneous
wind, wave, current and ice loading.
However, this accurate approach in the time domain is too cumbersome for fatigue
analyses during the early design stages. Therefore, simplified methods are developed.
The damage equivalent stress ranges from time domain simulation of the aerodynamic
response of an offshore wind energy converter in a calm sea are superimposed on the
results of a linear spectral analysis of the hydrodynamic response of the support
structure. Such separate calculations are acceptable if the aerodynamic damping of the
support structure is accounted for. Standard design tools from the communities of wind
energy and offshore technology can be applied conveniently for each task.
For the certification calculations still integrated, non-linear time domain simulation of the
simultaneous response is preferred, however, the scattering of wind and wave
parameters results in too many load cases from a practical viewpoint. A low number of
‘lumped’ load cases can be constructed by another approach.
The support structure, loaded by wind and waves, is emphasised in the field of dynamics
since the wind turbine is not significantly influenced by the hydrodynamic excitations.
The interrelation of design and dynamics is studied for seven examples for distinct
offshore wind energy converters. Three of them were actually constructed. Rated
capacities range between 500 kW and 3 MW and environmental conditions span from
a sheltered inland sea to exposed North Sea sites. The essentials of offshore specific
dynamics include the environmental description, eigenmodes and frequencies,
aerodynamic damping, uncertainty of the foundation behaviour and response under
fatigue and extreme conditions.
Recommendations for the dynamic analyses and design practice are compiled.
The design methodology, the analysis methods and the gained understanding of
dynamics are demonstrated during the stepwise development of the design for a 300
MW offshore wind farm.
Application during the European Opti-OWECS study and the realisation of the
Utgrunden offshore wind farm proved the benefits of an integrated consideration of both
dynamics and design process.
iii
CONTENTS
Summary
Notations and Coordinate Systems
vi
Acknowledgements
xii
1. General Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Design challenges for offshore wind farms . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Research objectives and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1
4
6
2. State of the Art of Offshore Wind Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 Offshore wind farm technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.1 Offshore design requirements
10
2.1.2 Components of the technology
11
2.1.3 Review and projection of technological development
18
2.2 Grid integration, energy supply and financing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Resources and Economics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Public acceptance, local environmental impact and politics . . . . . . . . . . . . . . . . . 22
2.5 Market activities, development targets and prospects . . . . . . . . . . . . . . . . . . . . . 25
2.6 Résumé and open questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3. Integrated Design Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1 Basic design considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Design practices in the parent technologies . . . . . . . . . . . . . . . . . . . . . . .
3.3 Presently applied design approaches for offshore wind farms . . . . . . . . .
3.3.1 Robust or traditional design approach
3.3.2 Parallel design approach
27
28
29
32
33
34
3.4 Innovation by the integrated design approach . . . . . . . . . . . . . . . . . . . . . . 38
3.4.1
3.4.2
3.4.3
3.4.4
3.4.5
General principle
Economic optimisation
Design for reliability, availability, maintainability and serviceability
Overall dynamics
Radical design for offshore requirements
38
40
41
43
43
3.5 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.6 First application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4. Description of the Offshore Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.1 Met-ocean parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.1.1 Wind
4.1.2 Waves and currents
4.1.3 Other parameters
53
60
68
4.2 Correlation of wind and wave parameters . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3 Soil conditions and auxiliary external effects . . . . . . . . . . . . . . . . . . . . . . . 73
iv
5. Modelling of Offshore Wind Energy Converters . . . . . . . . . . . . . . . . . . . . . . 77
5.1 Wind turbine modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.1.1
5.1.2
5.1.3
5.1.4
5.1.5
Non-homogeneities of the wind field
Rotor aerodynamics
Structural dynamics
Mechanical-electrical conversion system
Supervisory and power control
78
82
85
85
86
5.2 Support structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.2.1
5.2.2
5.2.3
5.2.4
5.2.5
5.2.6
General aspects of structural modelling
Piled structures
Gravity structures
Direct aerodynamic loading of the support structure
Hydrodynamic loading
Sea ice loading
88
90
95
97
98
103
5.3 Particular sub-system aspects that affect the entire system . . . . . . . . . . 104
6. Integrated, Modular, Non-Linear Time Domain Simulation . . . . . . . . . . . . 109
6.1. Integrated but modular dynamic model . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.1.1 Integrated model
6.1.2 Modular model
110
113
6.2. Time domain versus frequency domain approach . . . . . . . . . . . . . . . . . . . 116
6.2.1 Non-linearities
6.2.2 Fatigue analysis in time domain or frequency domain
116
117
6.3. Model reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.3.1 Reduction of degrees of freedom
6.3.2 Lumping of load cases for wind and wave fatigue
119
124
6.4. Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7. Bases for Simplified Fatigue Analysis Approaches . . . . . . . . . . . . . . . . . . 127
7.1 Separation of simultaneous response under wind and wave loading . . . 128
7.1.1 Independence of wind and waves
and dynamic interaction of responses
7.1.2 Contributions of wind and waves to the simultaneous response
7.1.3 Estimation of the aerodynamic damping
7.1.4 Empirical validation of separation of responses
128
132
134
140
7.2 Superposition of fatigue loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.2.1 Evaluation of different superposition strategies
7.2.2 Derivation of the superposition of equivalent fatigue loads
7.2.3 Empirical validation of superposition approaches
143
146
152
7.3 Empirical validation of superposition of separate fatigue analyses . . . . . 156
7.3.1 Short-term fatigue loads
7.3.2 Long-term fatigue loads
8.
Simplified Analysis of Aerodynamic and Hydrodynamic Fatigue . . . . . . .
8.1 Simplified analysis based on superposition of short-term fatigue loads .
8.2 Simplified analysis based on superposition of long-term fatigue loads . .
8.3 Establishment of load cases for integrated time domain simulations . . .
8.4 Parameter study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
156
158
161
162
170
174
179
v
9.
Interrelation of Dynamics and Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.1 Reference designs and sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2 Siting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.3 Dynamic characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.3.1 Natural frequencies and foundation influence
9.3.2 Excitation of tower and flapwise blade bending
9.3.3 Relevance of mode shapes
183
184
188
190
191
193
195
9.4 Aerodynamic and other sources of damping and inertia . . . . . . . . . . . . . 196
9.4.1 Aerodynamic damping
9.4.2 Hydrodynamic mass and damping
9.4.3 Soil damping
197
199
199
9.5 Foundation uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9.5.1 Evaluation measurements at the Lely wind farm
9.5.2 Sensitivity of eigenfrequencies
9.5.3 Improved design for wind farm Lely
201
202
203
9.6 Fatigue and extreme response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
9.6.1 Aerodynamic and hydrodynamic fatigue
9.6.2 Extreme response
204
211
9.7 Comparison of design drivers for the support structure . . . . . . . . . . . . . 215
10. Design Optimisation by Means of Tailored Dynamics . . . . . . . . . . . . . . . . 217
10.1 Feasibility study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
10.1.1 Treatment of dynamics and design
10.1.2 Preliminary design of the Opti-OWECS wind farm
220
225
10.2 Conceptual design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
10.2.2 Conceptual design of the Opti-OWECS wind farm
231
232
10.3 Detail design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
10.3.2 Detail design for the Opti-OWECS wind farm
238
239
11. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
A: Offshore Wind Energy Terminology
265
B: Technical Data of the Reference Sites and Designs
268
Samenvatting (in Dutch) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Zusammenfassung (in German) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
Curriculum Vitae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
vi
NOTATIONS AND COORDINATE SYSTEMS
Symbols
a
slope parameter in the definition of design turbulence intensity
[-]
axial, tangential induction factor
[-]
a, a’
a
annuity factor in energy cost calculation
[-]
nth vibration maxima
[m]
an
constant in the definition of the S-N curve
[-]
A
th
amplitude of i elementary wave
[m]
Ai
A
state matrix of state space model
input matrices of state space model
Bu , Bν
c
local chord length of rotor blade
[m]
[-]
cd , cl , cm aerodynamic drag, lift and moment coefficient
hydrodynamic drag, inertia coefficient
[-]
CD, CM
C
output matrix of state space model
Cu
undrained shear strength of clay
[kPa]
water depth
[m]
d
derivative of the lift coefficient with respect to angle of attack
[1/rad]
dcl / dα
dgen
generalized damping matrix
[Ns/m, Nms/rad]
equivalent viscous soil damping constant
[Ns/m]
deq
dhorz , dtilt , dvert horizontal, tilt, vertical foundation damping constant
[Ns/m, Nms/rad]
fatigue damage
[-]
D
D
outer diameter
[m]
DC
net decommissioning cost
[i]
[-]
D1 , D2 , D3 constants in the Dirlik expression
D
(physical) damping matrix
[Ns/m, Nms/rad]
Du , Dν
feedthrough matrices of state space model
expected value
E[ ]
levelised annual energy yield
[kWh/a]
Ey
f
vector of the external forces
[N, Nm]
cyclic frequency, general
[Hz]
f
first fore-aft bending eigenfrequency
[Hz]
f0
design value of first fore-aft bending eigenfrequency
[Hz]
fD
[Hz]
∆f , ∆fmin frequency increment of spectral discreatisation
rotor, blade passing frequency
[Hz]
fR , fb
[Hz]
fflap , ftower blade flapwise, tower fore-aft eigenfrequency
vector function for state derivative of wind turbine model
F
dF
hydrodynamic force per unit length
[N/m]
(disturbance of) aerodynamic thrust force
[N]
∆Fx , Fx
aerodynamic thrust force per unit length
[N/m]
dFx
g
gravitational acceleration
[m/s2]
dynamic shear modulus of soil
[MPa]
Gdyn
hub height with respect to terrain or mean sea level
[m]
hhub
vii
H
individual wave height
[m]
output vector function of wind turbine model
H
extreme wave height
[m]
Hmax
reduced wave height according to Germanisvher Lloyd
[m]
Hred
significant wave height
[m]
Hs
I
total investment cost
[i]
I
identity matrix
design turbulence intensity at 15 m/s mean wind speed
[-]
I15
IG
generator current
[A]
k
Weibull shape factor
[-]
modulus of initial subgrade reaction of soil
[MPa/m]
K
generalized stiffness matrix
[N/m, Nm/rad]
kgen
[N/m, Nm/rad]
khorz , ktilt , kvert horizontal, tilt, vertical foundation stiffness
Keulegan-Carpenter number
[-]
KC
K
(physical) stiffness matrix
[N/m, Nm/rad]
K1-value, stress range resulting in D = 1 for 2·106 cycles
[MPa]
K1
wave length
[m]
L
m
number of selected eigenmodes
[-]
generalized mass matrix
[kg, kgm2]
mgen
nth moment of the auto spectral density, n 0 { 0, 1, 2, .. }
[MPa2/sn]
mn
M
(physical) mass matrix
[kg, kgm2]
M
bending moment or torque
[Nm]
modal mass with respect to first fore-aft mode
[kg]
Mo
n
number of physical degrees of freedom
[-]
[-]
n(∆σj) , nj number of predicted stress cycles in j th bin
number of stress range bins
[-]
n∆σ
economic lifetime
[a]
ne
total number of stress cycles during process
[-]
ntotal
number of endured stress cycles, S-N curve
[-]
N(∆σ)
number of blades
[-]
Nb
number of endured cycles for the reference fatigue strength
NR
[-]
of the considered S-N curve, i.e. 2·106 cycles
p
probability
[-]
P
lateral load per unit length of pile
[N/m]
generalized coordinate or modal degree of freedom
[m, 1/rad]
q
OM
annual operation and maintenance cost
[i]
Q
constant in the Dirlik expression
[-]
spanwise blade coordinate
[m]
r
r
test discount rate in energy cost calculation
[-]
r
reference signal of controller in state space model
radius of foundation plate
[m]
R
R
[-]
outer blade radius, blade root radius
[m]
R, Rroot
controller matrix in state space model
Rc
std
standard deviation of stress time series
[MPa]
auto spectral density of stress response
[MPa2/s]
Sσσ
controller matrix in state space model
Sc
l th set of elementary load cases for lumping
Sl
t
time
[s]
viii
tice
T
TOM
T
Tp
Tprocess
Tz
u
u
thickness of ice sheet
individual wave period
total levelised annual ‘downline cost’
transformation matrix for static reduction
peak period, average duration between peaks
duration of process or load case
zero (up-)crossing period
controlled input vector of state space model
displacement
u&
structural velocity
u& w
ü
üw
UG
ν
V
V
Vave
Ve50
Vin , Vout
V1h , V3h
Vave
W
Wdiss
x
y
y
z0
Z
horizontal water particle velocity
structural acceleration
horizontal water particle acceleration
generator voltage
external input vector of state space model
instantaneous wind speed, in general
mean wind speed, in general
annual average wind speed
extreme gust wind speed with return period 50 year
mean wind speed for cut-in, cut-out
hourly mean, three hourly mean wind speed
annual average wind speed
apparent wind speed at airfoil section
dissipated energy during load cycle
system state vector, vector of physical degrees of freedom
lateral displacement of soil
output vector of state space model
roughness length for wind shear model
α
α
α
β
δ
ε50
γ
γ
γF , γM
γsub
γsat
Γ
ζ
ζ1P , ζNb·P
η- , η+
Θ
ΘtiltC.o.G
Θtop
incidence, angle of attack, seabed slope
seabed slope
wind shear exponent in power law
irregularity factor of broad banded signal, Tp / Tz
Vanmarcke bandwidth parameter
strain at 50% of peak stress for clay
peak enhancement factor of JONSWAP wave spectrum
parameter in the Dirlik expression
safety factor for loads, safety factor for material
submerged unit weight of soil
saturated unit weight of soil
gamma function
non-dimensional slope of sea bottom
exclusion range at rotor, blade passing frequency
lower, upper structural uncertainty range of fo
pitch angle
rotational inertia of structure with respect to centre of gravity
rotational inertia rotor and nacelle
[m]
[s]
[i]
[s]
[s]
[s]
[m, 1/rad]
[m/s]
[m/s]
[m/s2]
[m/s2]
[V]
[m/s]
[m/s]
[m/s]
[m/s]
[m/s]
[m/s]
[m/s]
[m/s]
[Nm]
[m]
[m]
[-]
[rad]
[rad]
[-]
[-]
[-]
[%o]
[-]
[-]
[-]
[kN/m3]
[kN/m3]
[-]
[-]
[-]
[-]
[rad]
[kgm2]
[kgm2]
ix
Λ
µ
ν
νD
ξ
ξo
ξaero
ξaero,prod
logarithmic damping decrement
inverse slope of the S-N curve
Poisson ratio of soil
correction factor for lumped sea state parameter
damping ratio as fraction of critical damping
damping ratio of fundamental mode
aerodynamic damping ratio
effective aerodynamic damping ratio during production
3.1416 ..
π
kair , kw , ksoil
air density, water density, undrained soil density
stress, general
σ
ice crushing strength
σip
∆σ
stress range
endured stress range for N cycles, inverse S-N curve
∆σ(N)
reference fatigue strength, detail category
∆σR
∆σeq
damage equivalent stress range associated with NR
vector of the j th eigenmode, i th component of nj
nj , ni,j
Φ
internal friction angle of soil
Φ
modal matrix
circular frequency, general
ω
rotational speed of the rotor, generator
Ω, ΩG
Ω
circular forcing frequency of soil
[-]
[-]
[-]
[-]
[-]
[-]
[-]
[-]
[-]
[kg/m3]
[MPa]
[MPa]
[Mpa]
[MPa]
[MPa]
[Mpa]
[m, rad]
[deg]
[m, rad]
[rad/s]
[rad/s]
[rad/s]
Superscripts
static reduction
^
amplitude
^
parameter of lumped load case
red
Subscripts
aerodynamic response
a
simultaneous or superimposed aerodynamic and hydrodynamic response
ah
controller
c
dynamic degree of freedom
dyn
equivalent
eq
generator
G
hydrodynamic response
h
,
,
,
general running indices
i j k l
master degree of freedom
M
related to fundamental eigenmode, e.g. damping, frequency
o
associated with response during parking, idle or failure state
parking
associated with response during production state
prod
rotor
R
slave degree of freedom
S
static degree of freedom
stat
associated with tower top
top
x
Abbreviations
API
BEM
CD
DAF
DiBt
GL
HAT
IEA
IEC
IFFT
JONSWAP
LAT
LPC
MIMO
MSL
NESS
O&M
Opti-OWECS
OWEC
OWECS
1P
Nb P
pdf
PI
PID
PM
PSD
RAMS
RP1, RP50
SDOF
SIMO
SWL
American Petroleum Institute
blade element momentum theory
Chart datum
dynamic amplification factor
Deutsches Institut für Bautechnik
Germanischer Lloyd
highest astronomical tide
International Energy Agency
International Electrotechnical Commission
Inverse Fast Fourier Transform
Joint European North Sea Wave Project
lowest astronomical tide
levelised production cost
multi input - multi output system
mean sea level
North European Storm Study
operation and maintenance
Structural and Economic Optimisation of Bottom-Mounted
Offshore Wind Energy Converters
offshore wind energy converter
offshore wind energy converter system, offshore wind farm
rotor frequency
blade passing frequency
probability density function
proportional-integral controller
proportional-integral-differential controller
Pierson-Moskowitz
auto power spectral density
reliability, availability, maintainability and serviceability
annual return period, return period 50 years
single degree of freedom
single input - single output system
still water level
xi
Coordinate systems
Unless explicitly stated, the orientation of the coordinate systems are according to
Figures 0.1 and 0.2.
XB
YB
ZB
MXB
MYB
Perpendicular to ZB, and pointing
out-of-plane towards the tower
Perpendicular to blade axis and shaft
axis, to give a right-handed co-ordinate
system
Radially along blade axis
lead-lag (in-plane) bending moment
flapwise (out-of-plane) bending moment
Figure 0.1: Blade coordinate system
XT
pointing South
pointing East
YT
vertically upwards
ZT
Origin at each support structure station.
Note that for steady-state calculations the wind,
wave, current and ice are deemed to approach
from the North.
Figure 0.2: Support structure coordinate system
xii
ACKNOWLEDGEMENTS
The preparation of a thesis can not be achieved without the aid and support of many
people and institutions.
First of all my wife Petra is acknowledged for her assistance throughout the long course
of this work. Without her patience and encouragement completion would never have
been reached. Her flexibility and open-minded attitude enabled me to experience
personal and scientific stimulus on our various stations between Hannover, Berlin,
Rotterdam, Delft and Rheine. Also Stephan and Matthias contributed indirectly by
motivating for the subject and by expressing their sympathy through creative illustrations
and Lego models of offshore windmills.
Many colleagues at the Section Wind Energy and the Workgroup Offshore Technology
at Delft University, at the Wind Energy Group of Technical University Berlin and in both
the wind energy and offshore industry also helped me in the good achievement of this
work.
My daily supervisor and friend Wim Bierbooms supported me in many ways since the
very first days at Delft and especially during the coordination of the Opti-OWECS study.
He is recognized on behalf of the entire staff of the Section Wind Energy.
Thanks are given to prof. dr. Jan B. Dragt for his scientific advice, personal
encouragement and vision when arranging the PhD research in 1993. Through many
discussions and fruitful comments prof. dr. ir. Jan H. Vugts acted as a sparring partner
for the offshore technology and balanced the research both in a scientific and applied
manner. Prof. Dr.-Ing. Robert Gasch at Berlin provided supervision at regular meetings
which enabled orientation beyond the daily business.
I am grateful to the participants of the Opti-OWECS project, namely Murray Ferguson
at Kværner Oil & Gas Ltd. and Tim Cockerill at the University of Sunderland, and
engineers at aerodyn Energiesysteme GmbH, ECN, Ecofys Energy and Environment,
Elsamprojekt A/S, Energie Noord West nv, Enron Wind GmbH, Garrad Hassan &
Partners, Kværner Turbin AB, NedWind bv, NEG-Micon A/S, Rambøll A/S, Risø
National Laboratory on providing scientific discussions and access to commercial data.
Many Msc. Students: Wolfgang Kurz, Albert Winnemuller, Silke Köhler, Tatjana Pauling,
Po Wen Cheng, Christian Schöntag, Andreas Manjock, Bernd Heins and Leon Hofland
contributed to the research.
The members of the Doctoral Examination Committee are also thanked for their
attention to this thesis.
The Opti-OWECS study was financed partly by grant JOR3-CT95-0087 from the NonNuclear Energy Programme of the European Commission. A late part of the research
was supported through the inter-faculty research program on decentralised production,
storage and distribution of renewable energy (DIOC 1) of Delft University of Technology.
1
General Introduction
CHAPTER 1
GENERAL INTRODUCTION
1.1 Design challenges for offshore wind farms
Rationale for offshore wind energy
With world attention focussed on the damaging impact of greenhouse gases, wind
energy is emerging as a serious contender for large-scale, cost-effective and clean
energy. The case for wind energy has been strengthened in recent years by the
significant reduction of its energy cost owing to more mature technology, market forces
and incentives. Currently the wind energy industry is subject to large growth rates
comparable to mobile telecommunication and other information technologies (Figure
1.1).
However, there is an important and increasing problem constraining further exploitation
of wind energy in parts of Europe. Limitations on land use in areas where the population
density is high are slowing down the installation of new wind farms. In some countries
in Northern Europe the public will no longer accept a significant increase of onshore
wind power capacity. The exploitation of the huge offshore wind resource, with
considerably less environmental impact than large onshore wind farms, will then become
crucial in providing for future green energy needs.
140,000
installed per year
cumulative development
25,000
120,000
100,000
20,000
80,000
15,000
60,000
10,000
40,000
5,000
20,000
2008
2006
2004
2002
2000
1998
1996
1994
0
1992
0
1990
MW per year
BTM Consult ApS
Year
Figure 1.1: Annual development of onshore and offshore wind energy
development up to 1999 and predicted growth of world capacity [1.1]
Cummulative MW
30,000
2
Chapter 1
Recent developments and perspectives
Between 1991 and 1997 the first small-scale offshore wind farms, rating two to five MW,
were installed in sheltered Northern European waters. They have clearly demonstrated
that offshore wind energy is feasible from a technological as well as an economic
viewpoint 1. Already in 1995 the ‘Study on Offshore Wind Energy in the EC’ showed the
large offshore wind resource [1.2]
These experiences and the generally accepted need for a higher contribution of
renewables, has significantly changed people’s attitude towards offshore wind energy
in the last few years. It sounds realistic when BTM Consult ApS predicts a 10 - 20%
contribution of offshore wind energy to the total new installed capacity of 9,000 MW by
the year 2004 [1.1].
The year 2000 marks the start of the commercial development by the installation of first
sea-based wind farms of intermediate scale and based upon the current generation of
megawatt wind turbines (Figure 1.2). For the year 2002 the commissioning of the first
large offshore wind farm with 160 MW capacity is planned in Denmark. For the more
distant future, plans have been developed for offshore wind farms in the gigawatt range
employing even larger multi-megawatt machines. It is imaginable that the installed
capacity of offshore wind power plant may eventually amount to several times that
installed on land [1.3].
Review of research
To date, research on offshore wind energy has concentrated on meteorological aspects
and both the design and dynamic loading of offshore wind energy converters. Until
recently most scientific publications have dealt mainly with national programmes for
long-term development, resource assessment, design studies or prototypes and
dynamics. Three conferences ‘Offshore Wind Energy in Mediterranean and Other
European Seas’ OWEMES 1994, 1997 and 2000 have highlighted the situation [1.4 1.6]. Inventories of the research requirements are presented in references [1.7] and
[1.8].
It is mainly to the credit of the European Non-Nuclear Energy Programmes JOULE and
THERMIE and the Danish and Dutch governments that first strategic projects have been
realised [1.2, 1.9 - 1.12].
Some understanding of design, dynamics and economics of offshore wind energy
conversion systems (OWECS) has been gained from various past design studies.
Several dedicated research projects are more important. For instance, Quarton and
Wastling [1.13] and Kühn [1.14] have developed analytical tools for design calculations
and highlighted the importance of the dynamics of the entire offshore wind energy
converter. Also the compilation of first design guidelines resulted from EU funded
research [1.15].
Nonetheless, we identified the following technical shortcomings; they form the
background of this thesis.
1
Giving a general overview of the state of the art of offshore wind energy is beyond the scope of this
introduction. For such information the reader is referred to the second chapter.
3
Figure 1.2: Utgrunden offshore wind farm commissioned in December 2000 as
example for offshore application of wind turbines in the current megawatt class
The economic performance of the demonstration plant is often poor due to its small
scale, its prototype nature and its lack of design optimisation. Recent design studies
have demonstrated the prospects of economies of scale and more mature technology
[1.9, 1.11].
The designs and design approaches applied have been governed by direct experience
either in onshore wind energy or offshore technology [1.16, 1.17]. Only very few studies
have developed solutions devoted particularly to offshore wind energy application [1.18 1.20].
Prototype installations were based upon a poor understanding of the dynamics of
offshore wind energy converters and its relation to physical and economic performance.
At the two Danish offshore wind farms Vindeby and Tunø Knob stiff, heavy and thus
expensive foundations were chosen whilst recent studies suggest advantages by softer
and lighter solutions and/or by tailored construction and installation techniques [1.19,
1.21]. At the first Dutch offshore wind farm Lely, simultaneous dynamic loading by wind
and wave was neglected in the design calculations [1.22] and a too low allowance for
uncertainties in soil stiffness led partly to unexpected dynamic characteristics [1.23].
Most of the previous design studies and research projects have examined the relation
between dynamics and design insufficiently. Moreover, dynamics have been
investigated at prescribed designs, rather than by demonstrating the opportunities of
design improvements by tailoring the dynamics.
4
Chapter 1
Some engineers have employed a quasi-static description of aerodynamic and
hydrodynamic loads only [1.24]. Other studies have considered fatigue in an overly
simplistic manner, i.e. either by taking into account only aerodynamic loads, [1.21, 1.25],
or using rough approximations of combined aerodynamic and hydrodynamic fatigue
[1.16, 1.26].
By contrast, analyses of integrated models of an offshore wind energy converter have
gained a basic understanding of the simultaneous response under wind and wave
excitation. In 1988, Oscar and Paez have shown the general effect of wind and wave
excitation on the response by Finite Element Analyses [1.27].
The attention was directed at the aerodynamic damping of the support structure by
Sinclair and Clayton who analysed a single degree of freedom model using a semiempirical approach [1.28].
Quarton and Wastling have extended two design tools for onshore wind energy
converters in the time domain and the frequency domain [1.13]. This facilitated the
investigation of three designs tested both in fatigue and extreme loading conditions. It
has been concluded that dynamic wave loading has little influence on the wind turbine
but that the support structure suffers dynamic wind and wave loads depending on,
among other parameters, its fundamental structural frequency.
Halfpenny [1.29] and Henderson [1.30] applied frequency domain methods to the
analysis of floating multi-unit offshore wind farms.
During the year 2000 common onshore wind turbine design tools as PHATAS, TURBU,
FLEX4 and HawC have been extended for offshore application; however, few results
from their application have been published yet [1.31 - 1.33].
Until now, analytical models of the dynamic behaviour of offshore wind energy
converters have not been validated by testing of actual plant.
The few existing structural measurements have dealt with dynamic loading due to
ambient and wake induced turbulence only [1.34], the eigenfrequencies [1.23] and
loading of monopile support structures [1.35]. More results are expected from ongoing
research projects under the Fourth European Framework [1.36].
1.2 Research objectives and methodology
Based upon the previous discussion we can state the main question of the present
Doctoral Thesis as:
How can analysis of the structural dynamics and a specific design approach,
improve the cost-efficiency and reliability of offshore wind energy conversion
systems?
Considering the three interrelated and overlapping fields of the dynamics of offshore
wind energy converters, the design and the optimisation of offshore wind farms, the
specific objectives of this study include:
•
Development and demonstration of a design methodology for offshore wind farms
Synthesis of analysis methods for the structural dynamics; the complexity of which
•
is adjusted to the different stages of the design process
•
Investigation of the particular dynamics of offshore wind energy converters and the
interrelation with the design
•
Demonstration of a design optimisation by means of tailored dynamics.
5
Dynamics
Interrelation of
Dynamics and
Design
(Chapter 9)
Simplified Analysis
of Aerodynamic
and Hydrodynamic
Fatigue
(Chapter 8)
Design
Optimisation by
Means of Tailored
Dynamics
(Chapter 10)
Economic
Optimisation
Integrated Design
Approach
(Chapter 3)
Design
Figure 1.3: Location of the research in the intersection of the three fields: dynamics,
design and economic optimisation
Figure 1.3 illustrates both the overlap between the three scientific fields and how
particular chapters of the thesis respond to the four objectives.
Development of the research project
The third of the above-mentioned goals, the investigation of the was stated at the
beginning of the research in 1992/1993. However, the three other objectives evolved
during the course as a consequence of the work itself and the rapid development of the
field of offshore wind energy.
The need for simplified, but not simplistic methods, suitable for application during the
early design stages of an actual design process was recognized after developing and
using a scientific analysis approach for the dynamics of OWEC which fits only to the final
design phase.
A broader consideration of the design problem followed from the coordination of the
European research project Opti-OWECS ‘Structural and Economic Optimisation of
Bottom-Mounted Offshore Wind Energy Converters’ (JOR3-CT95-0087) [1.11]. Offshore
wind energy was identified as a techno-economic rather than as a primarily technical
challenge; thus, design solutions require (economic) optimisation.
Furthermore, attention was directed to an integrated design approach considering
offshore wind farms as a whole and to coordinate knowledge in order to utilize the
valuable but partly contradictory experience from the parent technologies, i.e. (onshore)
wind energy, offshore technology and power management.
The Utgrunden project (Figure 1.2) offered the opportunity for us to apply parts of the
research to a real offshore wind farm during the years 1999 and 2000 [1.37].
6
Chapter 1
Applied methodology
Obviously different methodologies are appropriate to each of the three areas.
Dynamics were investigated by means of computational studies. Due to the lack of
existing offshore wind farms at exposed sites and of suitable measurements this was
the only possibility available. Nonetheless, studying two eigenfrequency measurements
at the Dutch Lely plant revealed important insight on monopile foundations. A major part
of the work concerned extending the onshore wind turbine design tool in the time domain
DUWECS (Delft University Wind Turbine Simulation) [1.38] by introducing the features
required for the description of the offshore environment and structural response. Semiempirical methods laid the foundation for the development of simplified fatigue analysis
approaches.
The integrated OWECS design approach was developed analytically and demonstrated
during the Opti-OWECS project. Furthermore, it was given a firm base by applying a
design methodology recently developed for large, complex civil engineering projects
[1.39].
Optimisation of actual designs was demonstrated by applying methodologies developed
in the parent technologies. It was beyond the scope of the study, and perhaps not
possible at all, to establish a mathematical model of the optimisation problem
appropriate to actual design practice. By contrast, emphasis was put on demonstrating
the required steps during the different design stages and by introducing approaches that
meet the requirements of these phases.
1.3 Outline of the thesis
Four chapters, each corresponding with one of the above-mentioned objectives (Figure
1.3), establish the core of this study. These major parts are embedded in a sequence
of further chapters as shown in Figure 1.4.
The framework of the research is outlined by this general introduction (Chapter 1).
Chapter 2 ‘State of the Art of Offshore Wind Energy’ provides a comprehensive overview
and can be read independently from the rest of the thesis. Readers who are not familiar
with the general background might use it as a starter before consuming the main course.
The integrated design approach is introduced in Chapter 3. It establishes a general
philosophy for innovative designs but can also be considered as a rationale for the
following elaboration.
The next chapters can be read in the actual sequence, however, they build up two
parallel development lines which may be considered independently corresponding to
personal taste and interest. Modelling of the environment and the OWEC itself and the
interrelation of dynamics and design are investigated in the first branch.
Secondly, appropriate approaches for the dynamic analysis are developed in Chapter
6 to 8.
Chapter 10 ‘Design Optimisation by Means of Tailored Dynamics’ presents the synthesis
of the two lines in the light of an integral treatment. Finally, the thesis rounds off with the
conclusions in Chapter 11.
Appendices with supporting information may serve as desert.
7
Special reference is given to Appendix A, which explains the terminology of this thesis.
It has been developed and used successfully during the Opti-OWECS project. In order
to avoid misunderstandings there are two conventions that must be appreciated. Firstly,
the acronym ‘OWECS’ (standing for offshore wind energy conversion system) and its
synonym ‘offshore wind farm’ describes the entire system, that is the wind turbines, the
support structures, the grid connection up to the public grid and the operation and
maintenance aspects. Secondly, ‘OWEC’ (offshore wind energy converter) is used to
refer to a single unit of an offshore wind farm comprising support structure (i.e. tower
and foundation) and the wind turbine (i.e. aero-mechanical-electrical conversion unit on
top of the tower).
(Chapter 1)
State of the Art of Offshore Wind Energy
(Chapter 2)
Integrated Design Approach
(Chapter 3)
Description of the
Offshore Environment
(Chapter 4)
Integrated, Modular, Non-Linear
Time Domain Simulation
(Chapter 6)
Modelling of
Offshore Wind Energy Converters
(Chapter 5)
Bases for Simplified Fatigue
Analysis Approaches
(Chapter 7)
Interrelation of
Dynamics and Design
(Chapter 9)
Simplified Analysis of Aerodynamic
and Hydrodynamic Fatigue
(Chapter 8)
Design Optimisation by
Means of Tailored Dynamics
(Chapter 10)
Conclusions
(Chapter 11)
Figure 1.4: Outline of the thesis
8
Chapter 1
9
CHAPTER 2
STATE OF THE ART OF OFFSHORE WIND ENERGY
The development of renewable energies has been identified as a key issue of the new
century and ambitious national and international plans for the reduction of greenhouse
gases have been announced. Offshore wind energy will be important at least for several
northern European countries since it has the potential for a considerable contribution
to the energy supply (Figure 2.1). Generally the prospects are assessed very positively
and the commercial development of a probably big market has commenced.
annual energy [TWh]
In this chapter it is tried to elaborate the state of the art of offshore wind energy with
respect to the dimension of the subject. We are dealing with five key areas which have
been defined with respect to scientific considerations: 1
•
Technology of offshore wind farms
•
Integration and financing of large-scale wind farms in the international energy
system
•
Resource assessment and economic potential
•
Social acceptance, environmental impact and politics
•
Market activities, development targets and prospects
Together with the résumé we formulate a number of open questions. Their solution will
be decisive for the future development.
3500
3000
2500
2000
1500
1000
500
0
water depth
< 10 m
< 20 m
< 30 m
< 10 km
< 20 km
< 30 km
< 40 km
distance from shore
(1989 European annual energy consumption 1727 TWh)
Figure 2.1: Estimated offshore wind energy potential in the European
Union (except Finland and Sweden) [2.1]
1
A similar approach is currently applied within the European project ‘Concerted Action on Offshore Wind
Energy in Europe’ (NNE5-1999-00562) which aims at a thorough assessment of all the critical issues
(continued...)
10
Chapter 2
2.1 Offshore wind farm technology
Since the early interest in offshore wind energy in the 1970’s, there have been a number
of significant advances in the understanding and the technology.
After briefly introducing four general design requirements, the individual elements of the
technology are presented. On this background we are able to review the development
so far and outline one possible future evolution of a mature technology in four
generations.
2.1.1
Offshore design requirements
The environmental impact and the demand of large power plants cause four distinct
design requirements of offshore wind farms:
•
Integrated design of offshore wind farms
More cost-efficient and reliable design can be achieved by an integrated design
approach considering the entire offshore wind farm as one system (Chapter 3).
Eventually optimum offshore design will differ from land-based design due to the
different distribution of the investment costs on the major subsystems and the greater
importance of operation and maintenance aspects.
•
Design for RAMS (Reliability, Availability, Maintainability, Serviceability)
Operation and maintenance (O&M) aspects are a main design driver for offshore wind
farms. Hindered access owing to bad weather after a fault results in longer down-time
and loss of production. Therefore, at sea the reliability must be even better than the
present good performance on land. Together with dedicated operation and maintenance
strategies and hardware, acceptable availability and reasonable O&M cost can be
achieved.
•
Smooth grid integration and controllability
Future large-scale offshore wind farms will be operated as power plants in the
international energy system. Fast controllability of the cluster as a whole as well as of
the individual turbines, for instance a power reduction by 80% within 2 s, are demanded.
Short power outages in the public high-voltage grid may not cause the entire wind farm
to drop out of production [2.3]. Among other considerations this is ruling-out robust
turbines with traditional stall control and almost constant rotor speed and to a lesser
extent fixed-speed, active-stall concepts. Extra cost and design effort are needed to
ensure high reliability of the more complex designs with variable-speed, variable-pitch
which are favourable for power management and load control.
•
Optimised offshore installation
Large costs but also a considerable potential for optimisation is related to installation of
offshore wind farms. Eventually installation procedures will be optimized until the entire
1
(...continued)
concerning the exploitation of offshore wind energy in Europe. The 17 participants from 13 countries
represent the utility sector, wind farm developers, engineering consultants, national agencies, research
institutes and universities. Conducted by a team of experts in the field, the work addresses the
technological, economic and socio-economic issues, and canvasses opinion from the financial, industrial,
environmental and political communities. The beginning of the two years project was initially proposed for
1998 but took place in the summer 2000 [2.2].
11
offshore wind energy converter including foundation is installed by only one to three
major offshore operations.
2.1.2
Components of the technology
Wind turbine
It is clear that the offshore environment poses particular requirements for wind turbines
which are not relevant to onshore machines. It is, however, also evident that in the
present state of offshore technology, the wind turbines above the height of wave impact
are little different from onshore machines.
Currently two approaches are followed to adapt the wind turbine design, i.e. marinisation
of robust and proven onshore solutions and the more radical design consideration of the
offshore requirements.
Figure 2.2 shows an example of the marinisation approach, the Enron Wind 1.5 offshore
realised at the Utgrunden wind farm in 2000 (Figures 1.2 and 2.5). Contenders of the
latter category are lightweight, high-tip-speed designs [2.4] and RAMS integrated
designs which were investigated for instance by the European Opti-OWECS study [2.5].
Ongoing rapid growth of the machine size and the requirement of proven technology
favour horizontal axis, upwind orientated wind turbines more or less similar to landbased solutions. Currently three-bladed designs are used while advanced concepts with
two blades might been applied at a later stage when a more mature level of the offshore
wind energy technology will be achieved. The simplified installation and maintenance
and the opportunity for higher tip-speed ratios offshore and associated lower drive train
loads favour two-bladed machines while the complicate dynamic behaviour and the high
fluctuating loads require sophisticated solutions.
Fixing for
temporary
hub crane
Direct inlet for
cooling air
Temporary crane
for generator
Improved
slip-ring
system
Vibration
monitoring
Improved
lubrication system
Trolley crane
Figure 2.2: Example of some marinisation features of the Enron Wind 1.5 offshore
used at the Utgrunden project
(Dimensions: length 8.8 m nacelle + 3.65 m hub, width 3.55 m, height 3.8 m)
12
Chapter 2
To date the minimum size of a single offshore turbine is the so-called megawatt class
with a rated power of 1.5 to 2 MW. However, large multi-megawatt units are under
development (Section 2.1.3).
As above mentioned a strong tendency towards variable-speed, variable-pitch machines
is observed and the further prospects of active stall turbines are questionable. With
respect to the drive train and generator concept the diversity is still present and even
increasing. The traditional concepts combine either a multi-stage gear box with a high
speed generator or no gear box (‘direct-drive’) with a low speed, electromagnetic multipole generator. Innovative designs have been proposed which match a single stage
gear with an intermediate speed generator (Figure 3.13) or use permanent magnet
generators and DC power transmission.
We come back to the wind turbine technology in Section 2.1.3 when presenting a
projection of the future development.
Support structure
With respect to bottom-mounted support structures Ferguson introduced a useful
classification by distinguishing their features with respect to three basic properties:
structural configuration, foundation type and installation principle [2.6]. Any structural
concept including mono-tower, braced or lattice tower can be reasonably combined with
either a piled and gravity foundation. From a logistic point of view it is, however,
beneficial to match piled designs with lifted installation and gravity foundations with
floated installation (Figure 2.3).2
Design optimisation with respect to local site conditions, batch size and the
particularities of offshore wind energy converters offers large cost reduction (Chapter
10). The dynamic characteristics of the support structure is of particular importance and
a compromise between the opposing requirements from the wind energy and offshore
communities has to be found. Wind turbine towers 3 in the megawatt class are slender
and flexible structures with a first eigenfrequency below 0.45 Hz. Common offshore
structures employ a significant higher natural frequency since dynamic wave loading is
strongly reduced for structural frequencies above approximately 0.4 Hz.
Floating offshore wind farms have been proposed for water depths beyond, say, 60 m
when bottom-mounted designs are not viable and the water depth is large enough to
enable efficient moorings. The highest proportion of capital costs is absorbed by the
moorings accounting, for instance, up to 28% in case of a proposed floating wind farm
with 1.4 MW turbines mounted on spar buoys [2.7]. Therefore innovative concepts as
the Multi Floating Unit Offshore Wind Farm (MUFOW) have been proposed which
reduce the wave induced response and both mooring and cable costs (Figure 2.4) [2.8].
Nonetheless they are subject to extra material, construction and construction-siteselection costs.
Recently Henderson drew the following sceptical résumé [2.9]. ‘All costing exercises into
floating wind parks have concluded that the costs will be at least twice that of ground
2
The floated installation of the tower and wind turbine on a lifted monopile is an interesting exception of
this classification.
3
Three design solutions exist for wind turbine towers depending on the ratio between the fundamental
eigenfrequency fO and either the rotor frequency fR or the blade passing frequency f b = N b @ f R : soft-soft,
i.e. fO < fR , soft-stiff, i.e. fR < fO < fb and stiff-stiff, i.e. fb < fO . In practice, stiff-stiff designs are no
longer built for the current generation of wind turbines.
13
Configuration
Piled & Lifted
Braced Tower / Lattice Tower
Gravity & Floated
Foundation & Installation
Mono Tower
Figure 2.3: Classification of bottom-mounted support structure types by
structural configuration, foundation and installation principle
Wind turbine
(fixed yaw position)
Walkway
Tapered hull
Water level
Mooring turret
Moorings
Power cable
Figure 2.4: Multi Unit Floating Offshore Wind Farm (MUFOW)
14
Chapter 2
Figure 2.5: Rotor installation at the Utgrunden wind farm with two jack-up barges
and two tug boats
based projects. Hence, if the concept is to progress beyond niche markets major
conceptual advances must be found. Combining with alternative energy sources, such
as wave or gas power is one possibility, as is generating an alternative higher value fuel,
such as hydrogen.’
Sometimes the re-use of decommissioned platforms of the oil and gas industry is
discussed. The viability of such isolated and small-scale solutions, if any, will most likely
be based upon the delay of the high dismantling cost of the original structure.
Installation, decommissioning and dismantling
Scarce equipment with high mobilisation costs, weather delays and safety requirements
make offshore operations about 5 to 10 times more expensive as on land. The logistics
required for such activities are illustrated by Figure 2.5 showing four of the seven
vessels and barges used for the installation of the Utgrunden wind farm. Therefore
especially installation and to a lesser extent decommissioning and dismantling are
design drivers of offshore wind farms. Minimisation of in-situ work, e.g. by reduction of
site preparation work or commissioning of the entirely equipped offshore wind energy
converters at the construction site will be important for the future success. Dismantling
and depositing of gravity foundations might reduce the prospects of such designs as
observed recently in the oil and gas industry concerning European and North-Atlantic
site. Complete removal of piles and submarine cables below the sea bottom, if
environmentally meaningful and legally required, might cause considerable costs.
15
Electrical transmission and grid connection
There are various possible approaches to the design and installation of systems for
power collection, transmission to shore and grid connection. Figure 2.6 gives the basic
options for the grid connection. No real technical restrictions are foreseen because
nowadays modular electronic components are available for a wide variety of
applications. The challenge lies apart from the economics in the integration into the
energy system which is discussed in Section 2.2.
The main choice which has to be made is between an AC or a DC connection to shore.
AC transmission involves high dielectric losses (the isolation material acts as a
capacitor); these losses are proportional to the cable length and the voltage. DC
transmission requires expensive converters. For short distances AC transmission is the
most cost effective option. The crossover point depends on the costs of the components
involved and is situated typically around 60 km.
HVDC (High Voltage DC) transmission systems have been increasingly used in recent
years to transport electricity from remote energy sources to the distribution grid. At
present the maximum capacity is 600 MW; 1,000 MW is expected to be feasible at
approximately the same cable cost per unit length by the year 2015 [2.10]. The progress
in power electronics has brought up Medium Voltage DC systems, also denoted as ‘DClight’, which might be interesting for medium size wind farms or the power collection
within clusters.
Connection to shore
AC
DC
[A1]
Collection within wind farm
G~
[C]
G~
G~
AC
~_
[A2]
G~
~~
G~
~~
[B]
DC
_
~
G~
[D]
G~
~_
G~
~_
_
~
G~
~_
G~
~_
_
~
Figure 2.6: Basic grid connection options
For the power collection there also has to be made a choice between AC and DC. The
first two options, A1 and A2, are the ones commonly used for onshore farms. The layout
according to option B has been used at some few onshore farms. Option C, AC coupling
of all wind turbines together with a DC connection to shore, may cause technical
problems with respect to achieving stable operation in the ‘AC island’.
DC power collection and connection to shore, option D, is elegant but meaningful only
if equipment for transformation of DC voltage for reasonable cost has become available.
16
Chapter 2
ABB in cooperation with ScanWind have announced the development of a machine
concept called ‘WindFarmer’ in this category.
Today generators operate with AC as the public grid does. This means that, where an
intermediate DC link is used, both AC/DC rectifiers and DC/AC inverters are required.
The converter stations consist, amongst other items, of thyristor switches. They have
to be placed in series because they can only switch a limited voltage (8kV). With
developments in semi-conductor technology, it is expected that the voltage which can
be switched by one thyristor will grow gradually. This means lower cost at equal power
and lower energy losses. IGBT (insulated gate bipolar transistor) switches can be used
alternatively.
Operation and maintenance (O&M)
The operation and maintenance requirements of offshore wind farms are crucially
important in the context of the overall cost effectiveness of offshore wind energy. Poor
performance and resulting low availability can offset economies of scale of large
offshore wind farms and multi-megawatt units. Two striking differences with onshore
wind farms are the largely reduced accessibility during bad weather conditions (wave
height, wind speed and visibility) and the costs of transport or lifting operations.
The demonstration offshore wind farms in sheltered waters show availabilities close to
the onshore situation. Unfortunately this positive experience is not representative for real
offshore conditions. Therefore in the scope of the Opti-OWECS project the operational
and maintenance behaviour of large wind farms at remote offshore sites was analysed
by Monte-Carlo simulations including random generation of failures and weather
conditions [2.11]. Figure 2.7 gives one qualitative example.
In the upper diagram the wind farm availability is shown as a function of both the fraction
of time when of the wind turbines are accessible and the reliability or mean time
between failure of the design.4 A site without restriction with respect to accessibility (i.e.
onshore) reaches 97% availability for a state of the art design increasing up to 99% for
an extremely reliable design under an adequate maintenance strategy. As weather
conditions get worse availability may fall down to 53% to 89% depending upon the
failure rate and maintenance approach.
The lower plot indicates the normalised energy cost for 20 years loan and 5% test
discount rate as a function of accessibility and reliability. The Opti-OWECS design
solution with 75% accessibility and improved reliability class is taken as 100%. The
assumption is made that the energy cost at 100% accessibility increases with increasing
reliability by approximately 9% due to the higher wind turbine investment. Harsher
weather conditions with better annual average wind speed will simultaneously lead to
a decreased accessibility and increased gross energy yield 5. With state of the art
reliability, cost values below 106% are not reachable since the extra gross energy yield
is more or less compensated by the lower availability. However, improved reliability will
4
Reliability data for the state of the art were derived from a large data base with turbines in the 500/600
kW class. The improved reliability level can be achieved with existing technology but extra expenses for
higher specifications, partially redundant components and dedicated O&M strategies. Innovative
technology and some years time for development and testing are required for the highly improved reliability
class.
5
Gross energy yield for 100% availability, the actual energy yield results from multiplication with the real
availability.
17
Availability
100%
Reliability
90%
highly improved
80%
70%
improved
60%
state-of-the-art
50%
100 %
(onshore)
80 %
(nearshore)
60 %
(offshore)
40 % (remote
offshore)
Accessibilty
140%
Energy cost
130%
Reliability
120%
state-of-the-art
110%
100%
improved
90%
80%
highly improved
70%
onshore
nearshore
100% accessibility
100% gross energy
offshore
80% accessibility 60% accessibility
130% gross energy 160% gross energy
Accessibilty &
gross energy yield
remote offshore
40% accessibility
190% gross energy
(at 100% availability)
Figure 2.7 Wind farm availability (upper) and energy cost (lower) as
function of the reliability of the wind turbine design for sites with different
accessibility and wind resources
(Energy cost for constant cost for support structure and grid connection)
find its pay back in terms of significantly reduced energy cost. Between 95% and 75%
accessibility the improved reliability class shows the lowest energy cost whilst for more
exposed sites the significant efforts for a highly improved reliability are cost effective.
Standards and regulations
The particularities of offshore wind energy limit the prospects for direct application of
existing onshore wind energy and offshore oil and gas expertise. The commercial
success of large international projects depends also upon sound technical criteria.
Establishment and harmonisation of specific standards has not progressed very far, yet.
In 1995 Germanischer Lloyd compiled first ‘Design Recommendations for Offshore
Wind Energy Conversion Systems’ [2.12] which are based upon a combination of
18
Chapter 2
onshore wind energy regulations, standards for fixed offshore installation and some
generic offshore wind energy knowledge [2.1]. Revision and possible extensions are
part of a recent European research project to be finished in 2002 [2.13]. In Denmark
design bases have been established which orient at the specific requirements of the first
national, large-scale projects [2.14].
The Working Group 3 of the Technical Committee 88 of the IEC has started the
preparation of a separate offshore extension of the IEC 61400 series to be in force by
the summer 2005.
2.1.3
Review and projection of technological development
The large success of wind energy during the last 15 years was possible due to the
gradual and evolutionary development of the technology and both market forces and
public incentives. The same is expected for offshore windpower, which stimulates us to
project four illustrative technology generations.
Twenty years between initial ideas and first demonstration projects
First ideas on electricity generation from the wind on the oceans date back to the early
1970's [2.15]. Here we can give only a brief overview of the development. Matthies et
al. [2.1] and Beurskens [2.16] provide more comprehensive information.
The initial feasibility studies on large offshore wind farms were undertaken in the USA,
the United Kingdom, Sweden, the Netherlands, Denmark and Germany between the
mid 1970's and the mid 1980's. One general reason why offshore wind power plants
were not installed in any of the countries concerned was that large-scale, commercial
wind turbines were not available at that time [2.17]. Moreover, some early studies
indicated too high cost in comparison with other energy sources as coal, gas and
nuclear.
Between the mid 1980's and early 1990's research was mainly focused on evaluating
the resources and the economics of offshore wind energy. In parallel with these studies,
onshore wind energy became of age evolving into a mature technology.
Prototype generation of the technology in the 500/600 kW class
Between 1990 and 1997 in Denmark, the Netherlands and Sweden the first
(semi-)offshore wind farms were installed and behaved well, particularly in view of their
prototype nature (Table 2.1). More or less standard machines of the 500/600 kW class
were placed in sheltered waters for demonstration purposes. The projects and the
converter units were too small to be competitive. The unexpectedly high energy yield of
the second Danish farm at Tunø Knob seems to have been something of a landmark
highlighting offshore wind energy as a feasible and potentially cost effective resource.
Research projects on Danish, Dutch and European level were initiated to start the
development of the required technology and tools for accurate prediction of the wind
climate. Utilities and other commercial enterprises from both within and outside the wind
energy community began to seriously consider commercial offshore wind farms.
It was also realised that in certain areas of Northern Europe the public acceptance of
onshore wind power projects may reach a saturation point within the foreseeable future.
In respond to this emerging barrier for further wind energy development, several national
energy agencies, e.g. in Denmark and the Netherlands, as well as organisations such
as Greenpeace drew up scenarios for the large scale exploitation of the offshore
resource with projections of installed capacity in the range of 4 to 10 GW per country
by the year 2020 to 2030 [2.18].
19
First commercial generation: Marinised megawatt class
The year 2000 marked the beginning of a new phase by installation of three small to
medium scale projects (Table 2.1) which concepts will also be used at the first largescale projects in the range of 100 to 160 MW to be realised in 2002 and 2003. The
megawatt class rated 1.5 to 2 MW power and using diameters up to 80 m establishes
the first generation of commercial offshore wind energy converters based upon
marinised onshore designs. This is a reasonable choice since the current offshore
market is too small for separate offshore designs but megawatt turbines are very well
suited owing to their onshore track record and being the largest commercially available
machines, yet.
Marinisation is related to various gradual design modifications as maintenance aids
enabling on-site repair of most components, improved nacelle climate and corrosion
protection (Figure 2.2) [2.19]. Some manufacturers introduced more significant changes
Table 2.1: Offshore wind farms realised until beginning of 2001
Location
Year
Power
[MW]
Rotor
diameter
[m]
Hub
height
(MSL)
[m]
Water
depth,
distance
from shore
Remarks
Nogersund,
Baltic, SE
1990
1 x 0.22
= 0.22
23
37.5
.7m
0.25 km
hybrid tripod with
gravity bases,
abandoned in
1998
(Figure 3.9)
Vindeby,
Baltic, DK
1991
11 x 0.45
= 4.95
35
37.5
3-5m
1.5 km
concrete caisson
Lely,
IJsselmeer, NL
1994
4 x 0.5
=2
40.77
40
5 - 10 m
0.75 km
driven monopile,
fresh water
(Front cover)
Tunø Knob,
Baltic, DK
1995
10 x 0.5
=5
39
43
3.1 - 4.7 m
6 km
concrete caisson
(Figure 3.6)
Dronten,
IJsselmeer, NL
1996
28 x 0.6
= 16.8
43
50
.5m
0.02 km
driven monopile,
fresh water
Bostigen Valar,
Baltic, SE
1998
5 x 0.5
= 2.5
37
.43
5.5 - 6.5 m
3 km
drilled monopile
Blyth Harbour,
North Sea, UK
2000
2x2
=4
66
58
8.5 m (MSL)
. 0.8 km
drilled monopile
Utgrunden,
Baltic, SE
2000
7 x 1.425
= 10
70.5
65
7.2 - 10 m
8 -12.5 km
driven monopile
(Figures 1.2 and
2.5)
Middelgrunden,
Baltic, DK
2000/
2001
20 x 2
= 40
76
64
3-6m
3 km
concrete caisson
20
Chapter 2
as introduction of full variable speed instead of limited speed variability or fixed speed.
Second commercial generation: Multi-megawatt designs for onshore and
offshore siting
The trend to even larger machines, especially on the German market, still continues.
The next turbine generation rated 2.5 to 3.5 MW with rotor diameters of 90 m and more
is currently on the drawing boards. The target markets onshore and offshore will
probably be served by site optimized variants. One 2.5 MW – 80 m onshore prototype
was erected at the beginning of the year 2000 and other larger machines are expected
in 2001 and early 2002 serving the second generation of offshore wind energy
converters (OWEC).
For instance, Enron Wind is currently developing a new platform in the 3 MW class with
different ratings for offshore and onshore siting. Many offshore requirements, e.g. higher
reliability of components, exchange of all major components with build-in cranage,
capsulation of the nacelle are considered already in the conceptual design. Many of
these offshore triggered features will be beneficial on land as well.
Third commercial generation: Dedicated multi-megawatt offshore designs
In our opinion the third OWEC generation rated 5 MW and larger with rotor diameters
of approximately 110 to 120 m will be applied more or less exclusively offshore. Siting
on land, apart from prototype installation, will be difficult among other reasons due to
the required infrastructure for transport and erection. The offshore wind farm proposals,
to be materialized during and after the second half of the present decade, are so large
that such dedicated designs will be fully viable and will be required. Furthermore
particular machine concepts, e.g. very high tip speed designs or two-bladed rotors,
might not be fully compatible with onshore requirements [2.4]. Also it might not be
optimal to design such large rotor blades for the higher turbulence, wind shear and more
pronounced directional changes onshore.
Some manufacturers claim to be able to jump directly from the megawatt class into the
4.5 - 5 MW range. However, time will show whether such an enormous step is
technically and commercially feasible or an intermediate stage, as planned for instance
in the Dutch ‘DOWEC’ project, is more realistic [2.20].
During the development of the second and third commercial OWEC generation an
integrated design approach considering the offshore design requirements will be
increasingly followed. Development of turbine size and technology will be most crucial
but similar innovations are expected and required with respect to support structure
design, installation and grid integration.
2.2 Grid integration, energy supply and financing
Integrating thousands of megawatt of offshore wind energy into the electricity system
presents a great challenge to the transmission grid. The present power and heat
production system in many countries is dominated by production from large-scale plants,
centrally located in areas with great population density, and to a lesser extent by
small-scale decentralized plants.
Massive development of offshore wind power means that electricity production from
wind turbines will lead to considerably greater periodic imbalances between production
21
and consumption than today. In addition the expected capacity variations from wind
farms are considerable. A balance must be struck between production, exchange with
countries abroad and consumption at any point in the day or night for reasons of
operational security in the electricity system. This means that other elements in the
electricity system will have to regulate the capacity variations caused by the integration
of large numbers of offshore wind farms.
Various types of countermoves to the periodic imbalance between production and
consumption of electricity are possible and are under development, for example:
• Shifting times of electricity production/consumption
• Pumped storage power stations
• Geographically spread production
• Converting electricity consumption to other power sources
• Plant adjustments and changed operations
• Development of innovative operation patterns by existing and new utilities
• Savings in heating buildings
• Foreign trade.
Reliable prediction of expected wind energy generation within the next 24 to 36 hours
will be important and can result in a significant price bonus [2.21, 2.22].
The financing of large offshore wind farms has to face different considerations as
commonly valid for onshore wind energy. Investment costs are generally higher due to
both the additional costs of offshore siting and the farm size. This together with the
generally longer economic lifetime and the larger uncertainties in the plant performance
results in a higher economic risk for the investor. The size of large onshore wind farms,
e.g. in Germany, Spain or the USA, is growing rapidly. Banks and financing bodies
seem to be prepared for large offshore projects in a due course.
2.3
Resources and Economics
Resources
It is clear from studies already undertaken that the offshore wind energy resource is
vast. In 1995 the ‘Study of Offshore Wind Energy in the EC’ [2.1] quantified the total
technically feasible European resources for bottom-mounted wind farms up to 40 km
offshore and 30 m water depth. Maps of the annual average wind speed for the coastal
European waters were computed and the potentially available areas for offshore wind
energy were summed up. Locations with too large water depth, distance to shore and
seabed slope were excluded as were areas for traffic lines, oil and gas exploitation,
nature conservation and military usage. The theoretical potential was estimated to
exceed the total annual electricity consumption (Figure 2.1). Some countries, namely
the United Kingdom, Ireland, Denmark and France are much richer in available energy
resources than others, e.g. Germany. Assuming a, say, 10% realisation of the total
potential results in a supply of 170%, 31%, 18% and 13.5% of the electricity demand in
the UK, Denmark, the Netherlands and France compared to 5% in Germany.
The extent and cost of exploiting the resource available to a particular country will
depend on the site conditions as well as the legal and economic constraints. During the
Opti-OWECS project, European maps with indicative energy cost were produced and
showed large resources at reasonable costs for the windy Northern European coastlines
and water depth up to, say, 25 m and moderate distance to shore [2.23].
22
Chapter 2
Economics
Various studies and offshore cost models suggest generally promising scenarios for
offshore wind energy developments. This represents a major change from an earlier
pessimistic view of offshore economics. It is now generally recognised that there is
compensation for the added cost of an offshore infrastructure in better annual average
wind speeds and lower wind turbulence offshore as compared with available land-based
sites that may service the same population area. Offshore, typical capacity factors, i.e.
the proportion of actual power generation to total potential generation, range 30 to 35%
whilst currently 23% is achieved on land [2.24].
Comparison of studies and actual realised projects in Figure 2.8 shows a dramatic and
continuous decrease in energy costs towards a level of other energy sources if equal
financial conditions are applied and site conditions are good. For instance, typical
energy costs based on 5% test discount rate and a repayment period of 20 years of
coal-fired and gas-fired plant range in the order of 3.7 to 5.5 ict/kWh and 3.1 to 4
ict/kWh, respectively [2.25]. In reality the loans for present wind energy projects are
shorter and conventional plants are amortised over a significantly longer period.
The cost breakdown of the energy cost of a 300 MW offshore wind farm proposed by
the Opti-OWECS project is given in Figure 2.9. Operation and maintenance costs have
a significant contribution. Optimum energy cost required high availability, which could
in turn only be achieved with a continuous and high O&M effort and permanent disposal
of heavy maintenance equipment.
The Opti-OWECS study identified the average wind speed of the site as the major driver
for the energy costs. While this may seem an unsurprising, indeed obvious conclusion,
it is difficult to find it in the published literature that more often refers to water depth and
distance from shore as design drivers. Other important environmental parameters are
the distance from the shore, the distance from electricity grid infrastructure, and the
water depth. Of the ‘controllable’ drivers, the most significant are the operation and
maintenance strategy adopted for the farm, along with the rated capacity of both
individual OWEC units and the whole wind farm. The design wave height and soil
conditions can have a smaller, although still significant influence on the farm economics.
This has a number of consequences. Excellent wind conditions are essential for the
economic viability of offshore wind farms. Estimation of the long term wind conditions
require great care if reliable energy cost predictions desired. Only if high availability is
guaranteed, the full potential of exposed sites can be exploited. It would be
economically worthwhile to invest a relatively large amount of capital in producing a
reliable design with a high energy output. For instance, an increase in gross energy
production or availability of 10% would bring an economic improvement even if it
requires 30% higher investment for the turbine.
2.4
Public acceptance, local environmental impact and politics
The first offshore wind farms with 4 to 11 wind turbines in the 500 kW class have been
regarded by the public as insignificant, test objects and the public interest has not been
of any importance. Today the public interest has changed considerably as the results
from the first wind farms are indicating a large potential for offshore wind energy. Typical
23
18
energy costs for 20 years loan,
5% test discount rate, no inflation,
1998 exchange rates,
bubble size propotional to farm capacity
RES, UK
energy costs [€ct / kWh]
16
14
UK Phase CII, UK
12
Horns Rev, DK
Bockstigen-Valar, SE
10
Vindeby, DK
Lely, NL
Middelgrunden, DK
Thyssen, DE
8
Nearshore, NL
Blekinge, SE
SK Power, DK
6
4
BMFT, DE
Opti-OWECS, NL
Scorby Sands, UK
Tuno Knob, DK
2
0
1990
1991
1992
studies
1993
1994
1995
projects
projects
1996
1997
1998
1999
2000
2001
reference year of study or project
Figure 2.8: Comparison of energy costs between recent studies and projects
Operation &
Maintenance
23%
Decommissioning
2.5 %
Wind turbine
34%
Project
management
1.5 %
Offshore grid
connection
15%
Support
structure &
OWEC instal.
24%
Figure 2.9: Contributions to energy costs of a 300 MW offshore wind farm in the
Dutch North Sea
24
Chapter 2
future farms are expected to comprise hundred wind turbines at least. This size has
caused concerns to the public and the environmentalists.
Public acceptance of onshore and offshore wind energy
Implementation of wind energy on land becomes difficult in some countries due to
scarce sites and/or poor acceptance. Under such conditions offshore siting is of
particular interest but has an ambivalent function.
Offshore windpower might be seen euphorically as the ultimate solution due to the large
potential and the lower environmental impact. Such an attitude might lead to a
frustration of the implementation onshore. For the manufacturers of the concerned
countries, e.g. the Netherlands, this would restrict their market to siting offshore or
abroad. Moreover it is questionable if the ambitious national goals for a sustainable
development by, among others, wind energy utilisation could be achieved in magnitude
and time ‘only’ by application of offshore wind energy.
Nonetheless, the public acceptance of offshore wind energy is not fully understood. In
the pioneer countries as Denmark and the Netherlands the attitude is generally quite
positive and the results of investigations of the environmental impact of the pilot plants
do not contradict this. Also Greenpeace, as important participant to the political
discussion, has formulated positive support.
At the same time however, environmentalists express their concern about the possible
effects of large-scale offshore wind farms. Also the visual impact of turbine clusters on
the ‘open seaside’ is sometimes feared and one is reminded on the NIMBY
phenomenon (‘Not-in-my-backyard’), well-known from many onshore projects.
Local environmental impact
A large-scale utilisation of offshore wind energy has very little known negative effect on
the local environment. Nonetheless, natural conservation areas as the Wadden Sea
National Park are excluded likewise as zones occupied by other uses.
Investigations at the small-scale pilot projects have not found significant influences on
migrating and resident seabirds, fish, sea mammals.
Environmental Impact Assessments (EIA) were completed without finding major
objections during the planning of the first four Danish 150 MW offshore projects and the
Dutch 100 MW Nearshore project [2.26, 2.27]. Topics covered by the Danish studies
included: hydrography, seabed conditions, water quality, marine biology, fish, birds and
marine mammals, visual impact, natural resources, marine archaeology, recreational
issues, planning issues, ships and navigation and fishery. The studies lasted about one
year and a half and will be continued by 2 to 3 years lasting investigations during and
after the construction phase. Similar investigations have been initiated in United
Kingdom, Ireland and Germany [2.28].
Eventually the experience with the first large projects available in 2002/2003 will
hopefully provide firm results which will be important for the future developments.
Politics
The necessary incentives to support a large utilisation of onshore and offshore wind
energy in the EU are completely dependent of decisions from the politicians. Certainty
about the conditions and duration of policy instruments, as a standard payment or a
market based system, are of high importance [2.29].Establishment of clear and fast
national planning rules and regulations is essential to limit the financial risks involved
25
in the planning process of a large project in the range of a quarter to half a billion Euro
even long before the construction phase.
Still a significant exploitation of offshore wind energy is a political decision as wind
energy cannot compete with fossil fuels if these are not judged by their social costs.
2.5
Market activities, development targets and prospects
Market activities and development targets
Commercial developments of the offshore resources have started in nearly all coastal
countries of the European Union. Here only a snapshot of some of the activities at the
time of writing can be given. Commercial interests and uncertainties about the outcome
and duration of the permission procedure complicate any prediction.
Most firm are five Danish projects of in total 750 MW to be realised before the year
2008. They are part of the long-term and well-coordinated process driven by decisions
of the government aiming at 4,000 MW of installed offshore capacity by the year 2030
corresponding together with the onshore generation to about 50% of the national
electricity consumption.
At least each one project of 100 MW is expected in Belgium and the Netherlands for
2002/2003 and 2003/2004. The fast realisation of the Dutch 100 MW demonstration
project ‘Nearshore Wind Park’ will be decisive for the further national development of
wind energy and the industrial opportunities. On the longer term 1,250 MW offshore
wind power are announced by the Dutch government for the year 2020. It is also
imaginable that a more rapid scenario is feasible [2.18].
A couple of projects of different size are under preparation in Sweden but low power
purchase prices are a potential problem.
In the United Kingdom two projects have received old NFFO 4 grants but realisation is
unclear at the moment. Up to ten other projects in territorial waters are in the planning
stage but permits are uncertain as learnt from the frustrating situation onshore. The
British Wind Energy Association considers it unlikely that any of these will be actually
be built until 2005 [2.30].
The situation in Germany has improved since the new Electricity Feed Law (EFL)
applies also within the territorial waters and for plants commissioned before the end of
the year 2005. Nonetheless, no prototype project is foreseen in a due course. So, wind
farm developers have announced big claims and started planning activities for more
than 15 projects in the range from 9 to 1,500 MW [2.31]. The size of the projects is
driven by large water depths of 20 to 35 m and long distances to shore of 15 to 60 km
which are required with respect to the conservation areas and the naval traffic lines.
Realisation will depend among other aspects on the complicated permit procedures,
financing and the availability of mature technology tested at foreign sites.
Ireland, France, Spain and Finland show activities as well. Some Mediterranean
countries, e.g. Italy, have started exploitation of their resources and feasibility studies.
Potential interest outside Europe is seen in Japan where for instance the feasibility of
a renewable energy island combining various energy sources was studied. On the
longer term also other countries with shallow waters and good wind resources, e.g. India
or Bermuda, could be interested.
Exploitation of offshore wind energy is in line with the EU White Paper ‘Renewable
Sources of Energy’, where the specific target for wind power is set at 40,000 MW
installed capacity by 2010, 25% of which will be delivered by 10,000 MW capacity of
26
Chapter 2
large wind farms, most of them will be offshore wind farms [2.32]. The latest targets of
the European Wind Energy Association (EWEA) formulated in October 2000 consider
the ongoing rapid growth and exceed the old target which coincided with the EU White
Paper. 60 GW installed wind power including 5 GW offshore are projected for the year
2010 increasing to 150 GW including 50 GW offshore by the year 2020 [2.24].
Benefits to the environment
Investigations from Denmark show that wind energy is one of the cheapest ways of CO2
reduction compared to several other renewable energy sources [2.33]. The emissions
and waste avoided by 1 kWh of wind generated electricity depends on the national
energy mix and other factors. Taking Denmark as example each kWh wind electricity
avoids 850 g of CO2, 2.9 g of SO2, 2.6 g of NOx, 55 g of slag and ashes and 0.1 g dust
[2.25]. According to the British Wind Energy Association each MW of installed onshore
wind energy capacity supplies electricity for about 600 average households.
Benefits to the employment
There are recent signs of the established offshore engineering industry, energy
conglomerates and industrial combines beginning to give serious attention to wind
energy projects in particular offshore. The development of a market for large scale
offshore wind farms will bring new business to the wind turbine and offshore engineering
industries, shipyards and shipping branches of Europe, creating employment for the
associated work forces. Under present European market conditions 1 MW of onshore
wind power installed creates jobs for 15 to 19 people [2.25]. The number will likely be
higher for offshore projects.
2.6
Résumé and open questions
Offshore wind energy, a benign exploitation of the natural resources of the continental
shelf, undergoes a process of rapid development in several European countries.
Encouraging experience with the first large-scale project to be realised within the next
two years will decide upon the future of an annual market in the billion Euro range.
With respect to such prospects a number of questions, essential for the future
development of (offshore) wind energy, arises:
• How can the economics and reliability of large offshore wind farms be improved?
• Which technology is required for such an enormous application at exposed sites?
• What kind of approaches are necessitated for design and project realisation?
• Are current certification rules and standards suited for offshore developments?
• How to integrate large wind farms in the international energy supply system?
• What is the attitude of the financial world and the decision makers?
• Will offshore wind energy be cost effective only with bottom-mounted concepts and
mainly for a number of northern European countries?
• What are the technological and socio-economic conditions needed to encourage the
future development?
• How to avoid conflicts with environmental and other interests?
• How to achieve a beneficial development for onshore and offshore wind energy?
The rest of this thesis is related to the first four of these questions.
27
CHAPTER 3
INTEGRATED DESIGN APPROACH
Optimal design is facilitated by considering an offshore wind farm on an integral system
level (Figure 3.1). The so-called ‘integrated design approach’ is introduced to achieve
more cost-efficient and reliable offshore wind energy conversion systems (OWECS).
Other suitable design methodologies undoubtedly exist, but an integrated design
approach will most likely be nearer to the optimum in both technical and economic
sense.
Prior to the description of the novel approach, attention is focussed on the life cycle and
the design objectives (Section 3.1). Next, the design practices in wind energy
technology, offshore technology and power management (Section 3.2) and design
approaches applied for offshore wind farms so far (Section 3.3) are analysed. After this
preparation the integrated design approach is presented in Section 3.4. Sections 3.5
and 3.6 summarise a possible implementation and first experience on the application
in a near-design situation of the Opti-OWECS study.
More comprehensive information about the new approach can be found in the first
volume of the final report of the Opti-OWECS study [3.1].
wind turbine
aerodynamics
site
support
structure
hydrodynamics
farm layout
1
economics
grid connection
operation & maintenance
Figure 3.1: Sub-system and aspects of an offshore wind energy conversion system
(OWECS)
28
Chapter 3
3.1 Basic design considerations
Life cycle of offshore wind farms
Like most other products, the life cycle costs of an offshore wind farm are largely
determined during the design process which accounts only for a relatively small
proportion of the total investment. No fuel cost exist for this renewable energy source,
therefore the life cycle costs are mainly built up of construction and installation costs
initially and operation and maintenance cost during the operation phase, both closely
dependent on the designer’s choices.
Design: Project identification
Feasibility study, conceptual design
Structural design, specification
Site survey: desk top
Site survey: in-situ
Permissioning, financing, marketing
Tendering & Contracting
Implementation: Construction
Installation
Commissioning
Exploitation (operation & maintenance)
Post-Exploit.: Refurbishment (optional)
Dismantling
Time
????
????
optional
1 year 1 year 1 year 1 year 1 year 15 - 20 years
1 year 1 year 1 year
Figure 3.2: Qualitative life cycle of an offshore wind farm
Figure 3.2 shows a qualitative life cycle. Between the initial plan and the commissioning
typically up to five or more years may elapse for the three main steps: initiation, planning
(i.e. environmental investigation, permission procedure, design, financing and tendering)
and building (i.e. construction, installation and commissioning). Gaining all required
permits is a potential source of delays. However, also site surveys may require a
considerable period because in-situ soil investigations have to be done for each turbine
location and local wind speed measurements for at least one year are needed for a
reasonable prediction of the expected energy yield.
The operation phase may typically continue for 15 to 20 years. From an economic point
of view even longer periods would be advantageous. However, this could only be
achieved after a refurbishment of the wind turbines and other components.
Dismantling of offshore wind energy converters, especially if founded on piled steel
structures, is expected to be relatively unproblematic. Removing and depositing of
ballast of gravity structures may require extra considerations.
All tasks involving heavy works on site such as site measurements, installation,
commissioning and dismantling have a strong seasonal dependancy. If not completed
within one summer period it is impossible to continue at most sites before late spring of
the next year.
Objectives for optimal design
An offshore wind farm designed in an optimum manner will provide energy in the most
reliable and cost-efficient way over the projected lifetime.
29
Four design objectives for the entire system can be stated:
•
Tailored distribution of investment and operation and maintenance (O&M) costs
over the entire offshore wind farm and its lifetime.
The economics of the complete plant has to be balanced with respect to the overall
operational goal, which could be the achievement of the minimum cost of energy,
the delivery of a certain power quality and persistence or a combination of both.
Such goals cannot be reached by optimisation of separate sub-systems alone.
•
High reliability of the OWECS as a whole and of essential sub-systems.
A failure of a major sub-system can result in a loss of production for a long period.
For instance, many converter units could be out-of-operation due to a design
mistake or power cut-off in the grid connection system. Such a failure together with
unfavourably high repair costs might jeopardise the entire project.
•
Adaptation to economy of scale and partial redundancy of single converter units.
A typical large offshore wind farm, as regarded feasible this decade, comprises
between 50 and 100 offshore wind energy converters rated approximately 2 to 5
MW each. For such a batch size, design should be optimised with respect to the
particular environmental and economic conditions. Consideration of a partial
redundancy of the single OWEC units with respect to the production of the entire
wind farm might be worthwhile, for instance, within the operation and maintenance
strategy or when determining a design probability of failure.
•
Symbioses of experience from wind energy technology, offshore technology and
power management.
Optimal design is not achieved by the ‘blind’ application of common design
practices from the involved disciplines and therefore requires a joint solution.
3.2 Design practices in the parent technologies
The design of offshore wind farms should be based, of course, on the accumulated
experience of the wind energy technology, offshore technology and power management
communities. However, one has to be aware of the different backgrounds and
sometimes even contradictory principles in the disciplines which are summarised briefly.
Design of wind energy converters and wind farms
Wind turbine engineering has undergone a rapid development in the last ten years and
is now ‘coming of age’ [3.2]. During this period the average rating of commercial wind
energy converters increased by a factor of ten due to both better understanding of the
technology and demands of the market. The price per installed capacity and the price
per annual produced energy lowered significantly.
The increasing dimension of large turbines lead to high structural flexibility and potential
sources of coupled vibrations. Specific knowledge has been developed to achieve safe
solutions under such conditions e.g. by increased material or active damping, machine
supervision and advanced control. Soft tower designs are successfully applied for
reduction of weight, dynamic loads and hence cost.
30
Chapter 3
100%
Installation
80%
Transportation
60%
40%
Procurement
20%
Design
0%
Onshore Wind Farms Small Offshore Oil &
Gas Platforms
first eigenfrequency [Hz]
Figure 3.3: Indicative cost breakdown of onshore wind farms in
Germany [3.3] and small offshore oil and gas platforms [3.4]
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Onshore Wind Energy Converters
Soft-stiff
Soft-soft
30
40
50
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
60
70
hub height [m]
80
90
Offshore Oil & Gas Platforms
0
100
200
300
water depth [m]
400
Figure 3.4: Comparison of fundamental eigenfrequency between
wind energy converters and offshore oil and gas platforms
31
Most manufacturers produce between two and four different standard machines.
Modifications of these base line machines, also denoted as ‘platforms’, to tailor them to
specific location classes, take account of mainly two site parameters only. The annual
average wind speed influences the fatigue loading and determines the rotor diameter
and hub height while the extreme wind speed is an important factor in the strength
analysis. The atmospheric turbulence intensity is in most cases prescribed by the
applied standard. Sometimes site specific loads are calculated when the standardised
conditions deviate too much from the actually expected. Differences in onshore soil
conditions are compensated for by making minor variations in the foundation design
rather than by using different towers.
Wind farm design is mainly driven by the selection of a suitable machine type and size
from the range of standard designs, arrangement of the turbines under sufficient
distance to each other to limit the wind farm induced turbulence, compatibility with the
existing grid infrastructure and noise limitations.
Operation and maintenance aspects are important in order to ensure a sufficient lifetime
and minimise repair costs.
The cost breakdown of typical onshore wind farms (Figure 3.3) is dominated by the
procurement cost. Transport and installation cost are moderate but become more
important for large turbines and remote sites. The contribution of engineering costs is
low owing to series effects.
Design of offshore structures
There is a long experience in offshore technology with the design of large and unique
fixed structures for the petroleum industries which are built ‘fit for purpose’ with respect
to their particular site and function. In contrast to wind turbine design, the influence of
dynamic response under wave loading is generally limited by relatively high structural
stiffness in relation to the total height of the installation (Figure 3.4). The wave excitation
range extends between f = 0.04 Hz and f = 0.5 to 1.0 Hz. As a rule-of-thumb dynamic
wave loading becomes important for natural frequencies below 0.4 Hz in shallow waters.
So, from the over 7000 fixed offshore installations world-wide only a small number of
platforms in deeper water have eigenfrequencies below 0.3 Hz.
Although fatigue is important, generally it takes second place to the dominant extreme
event loading conditions.
Transportation and installation issues are often main design drivers since these costs
can be even higher than those for the manufacturing of the structure onshore (Figure
3.3). Reduction, and where possible, elimination of underwater inspection and
maintenance is essential because of the difficult access and the high cost associated
with these operations offshore. Other important design aspects concern the safety of
personnel working on or travelling to the platforms, environmental impact and
removal/dismantling.
Design of conventional power plants
Large coal, oil, gas or nuclear fired power plants are built since the introduction of
distribution lines for electrical power at the beginning of the last century. The power
demand pattern (base load or peak load) and the power type (electricity or both
electricity and heat) are the most important design requirements. Plants are controlled
according to the instantaneous and the short-term predicted power demand and are
integrated in the national and international network. A mixture of locally distributed
32
Chapter 3
stations with different response times and a certain overcapacity ensures a reliable
supply.
With the exception of hydropower plants, the site is chosen close to the consumption
centre or where fuel and cooling water can easily be supplied. Also, environmental
aspects are important.
Power plants are designed often in a modular way. So the total capacity is provided from
a combination of a number of standard power units more or less out of series
production.
Several reasons make the grid integration of large wind farms a new challenge to the
power management and wind energy communities. Site selection is now governed more
by the natural resources and legal constraints rather than the location of the
consumption and the available infrastructure. The short-term and long-term variation of
the supply of wind energy has to be matched with the fluctuation of the energy demand.
This requires wind energy prediction tools and both accurate and fast controllability of
the entire wind farms as well as of the individual converters [3.5]. Furthermore, the
economical value of wind generated electricity can be significantly lower if spot market
prices instead of fixed feed-in rates are considered [3.6].
3.3 Presently applied design approaches for offshore wind farms
The first prototype offshore wind farms and many of the studies and projects employed
design approaches, which are denoted in this thesis as either the ‘robust or traditional’
or the ‘parallel design approach’.
The organisation within the design approaches is explained by the terminology and
hierarchical diagrams as developed recently for the design and construction of complex
civil engineering systems [3.7]. Such a treatment has been successfully applied, for
instance, on projects as the Storm Surge Barrier in the Nieuwe Waterweg near
Rotterdam and the Ekofisk Protective Barrier in Norway. With some adjustments it
seems also suitable for offshore wind farms.1
On the one hand it is useful to decompose the system with respect to its variables or
requirements in order to organise the design of such systems. The relation between the
variables (upper left part of Figure 3.5) represents the total friction between solution and
problem.
On the other hand structural control is simplified by a decomposition with respect to subsystems or design clusters and their elements (upper right part of Figure 3.5). In theory
the boundaries of the sub-systems should be governed by maximum internal and
minimum external (inter-)relations. However, often practical considerations of involved
disciplines, organisations or materials are of equal or even larger importance.
Simultaneous goal and structural coordination is facilitated by merging of the two
decompositions in an improved composite constructive diagram (lower part of Figure
1
An offshore wind farm comprises several elements that are not typical for most civil engineering projects,
e.g. a ‘machinery part‘ supported by a ‘building/structure part’, strong interaction between both elements
and a large number of identical units. Nonetheless, it seems worthwhile to apply here the experience
gained in the design and construction of complex civil engineering projects.
33
Conceptual solution
Requirements
+
design program
Structure
diagram of concepts
Coordination
System
=
Organisation
members
Client
Goals /
aspect systems
Project manager
Sub-systems
Cluster leaders
Elements
Engineers
improved composite constructive
Figure 3.5: Decomposition of a complex civil engineering system (upper part) and
established system for goal control and structural control (lower part) [3.7]
3.5). The clustering of requirements yields aspect systems2 for goal control whilst the
clustering of elements results in sub-systems for structural control. Without structural
control on a sub-system level no goal control on the system level is possible.
The different levels of the composite diagram are related to certain part systems of the
system, i.e. system, goals, aspect systems, sub-systems and elements, noted at the left
side of the diagram. At the right side certain members of the project team are assigned
to the different levels.
3.3.1
Robust or traditional design approach
The second Danish offshore wind farm at Tunø Knob (1995) [3.8] is a good example of
the robust or traditional design approach. To a lesser extent, this holds also for the
earlier farm at Vindeby (1991) [3.9]. In both cases the attribute ‘traditional’ refers to an
onshore design approach applied to the offshore environment.
2
An aspect system refers to issues or topics of a system which are a cluster of relations between elements
and/or individuals. Examples of aspect systems of a civil engineering system are: weight, strength,
stiffness, maintenance, costs, etc. A suitable set of aspect systems represents the behaviour of the system
and can be used for goal control on effectiveness. Therefore aspect systems should be chosen as specific,
as independent, as equally important and as quantifiable as possible.
34
Chapter 3
The main objectives of both projects were the demonstration and the investigation of
environmental effects. Therefore, and owing to the small project scale, more or less
standard onshore wind turbines were used. At Tunø Knob, wind turbines and towers
were installed by a large floating crane but in a similar manner as onshore, i.e. by
separate lifting of tower, nacelle and rotor (Figure 3.6). Stiff gravity caissons are
countering the design ice loads and are acting as artificial islands. Reduction of
operation and maintenance costs was aimed for by well-proven onshore wind turbine
design, marinised by features as improved corrosion protection, air-tight nacelle and
built-in lifting facilities.
The design process (Figure 3.7) is organised in two main clusters according to the
parent technologies. Wind turbine, tower, farm layout and grid connection design belong
to the wind energy technology while the offshore technology deals with foundation,
subsea cables and marine operations. Important aspects affecting the overall
performance as installation or operation and maintenance are treated separately by the
two technologies as elements of sub-systems rather than as separate sub-systems or
as aspect systems.
Clearly this approach fits the objectives of the prototype projects but is not suitable for
the proposed large scale offshore wind farms.
3.3.2
Parallel design approach
The parallel design approach considers the design requirements of offshore wind farms
separately for the main sub-systems wind turbine, support structure, grid connection and
operation and maintenance (Figure 3.8).
The system goals are decomposed in aspect systems specific for the different subsystems. For instance, costs as an important goal are controlled for every sub-system
but not on the system level. Thus, the economic performance is limited to the
summation of the optimised costs of the sub-systems. As is the case with the robust or
traditional approach, installation aspects and operation and maintenance aspects are
treated only on the sub-systems level.
The Dutch pilot project Lely in the IJsselmeer (1994) [3.10], the Swedish Nogersund
onshore wind energy converter (1990) [3.11], the Danish Middelgrunden wind farm
(2000/2001) [3.12] and the tenders for the first two 160 MW projects in Denmark
(2000/2001) can be mentioned as examples.
At Lely novel design solutions were applied by monopile foundations and a cable laying
technique reducing the use of a cable laying ship. Both innovations were a success, in
principle. Nonetheless, the system aspects have not been fully considered, for instance,
in the investigation of the overall dynamics (Section 9.5.1) [3.13]. Again, this could be
accepted because of the sheltered site and the demonstration nature of the project.
The installation procedure at Nogersund was most spectacular due to the radical
minimisation of in-situ work. The entire offshore wind energy converter was fully
assembled and commissioned in a dry dock before it was floated with auxiliary ballast
and towed to the final destination (Figure 3.9).
The tendering and construction of the first large Danish project at Middelgrunden project
with 40 MW installed capacity was separated into three parts for wind turbine and tower,
foundation and grid connection. Turbine and foundation were designed independently
35
Figure 3.6: Piecemeal installation
of Tunø Knob wind farm as example
of the robust or traditional design
approach
(Pictures by W. Steche,
Windkraft Journal 4/95)
Demonstration plant
Political order,
Client
System
Prototype
Study of
environmental effects
Limited risk
Project
manager
Goals
Subsystems
Elements
Wind Energy
Technology
Offshore
Technology
Wind turbine
(modified)
Foundation design
& construction
Tower
(onshore design)
Ancillary design
Farm layout
Installation
(wind turbine & tower)
Grid connection
design
Cluster
leaders
Engineers
Foundation
installation
Submarine cable
laying
Foundation and
cable maintenance
Operation &
maintenance
(wind turbine & tower)
Figure 3.7: Controlling of the robust or traditional design approach
Elements
Subsystems
Aspect
systems
Goals
System
Costs
O&M
Installation
(support structure
or OWEC)
Operation &
maintenance
(wind turbine,
farm)
Grid connection
maintenance
Farm layout
Cable laying
Switch gear,
transformer, etc.
Cable design
Grid connection
and farm layout
Reliability
Figure 3.8: Controlling of the parallel design approach
Support structure
maintenance
Ancillary design
Foundation
& tower
Support structure
Costs
Lifetime
Strength &
Costs Stiffness Installation
Energy yield
(Wind turbine
installation)
Wind turbine
(modified)
Wind turbine
Marinisation
Costs of energy
Demonstration plant
or (near) commercial plant
Engineers
Cluster
leaders
Project
manager
Client
Order
36
Chapter 3
37
Figure 3.9: Commissioning of the complete Nogersund offshore wind energy
converter in the dry dock as example of the parallel design approach
and after the selection of suppliers for both sub-systems, modifications of the foundation
were required. Also, it was more difficult to optimise the installation process for
foundations, cables and machines.
The main sub-systems of the first two 160 MW projects in Denmark will also be
tendered separately, but currently procedures are sought to optimise the entire system.
The parallel approach is typical for demonstration projects and first commercial plants
and might have some advantages if the client wishes to select separately the supplier
for the sub-systems rather than to design or tender the project as a whole. Resources
and extra time are required for the communication between the design clusters. Good
matching of sub-system solutions and fulfilment of the system goals might be achieved
only after a number of iterations.
As a consequence of the robust or traditional approach or the parallel approach it might
happen that the certification is done with respect to both a wind turbine standard and
standard for offshore structures. Such a cumbersome procedure involves conservatism
in the wind turbine loads if for instance the lower atmospheric turbulence offshore is not
considered.
38
Chapter 3
3.4 Innovation by the integrated design approach
3.4.1
General principle
The integrated design approach is treating the offshore wind farm as an entire system,
i.e. the offshore wind energy conversion system (OWECS). Interactions between subsystems are considered, therefore, as complete and practical as possible. So the
design solution is governed by overall criteria, the aspect systems. Examples for such
criteria include levelised production costs, dynamics of the entire system and OWECS
availability. Although such an approach will most likely produce an novel design solution,
the evolutionary nature of technical progress and the commercial risks inherent to
innovative solutions must be kept in mind. In particular it would be a serious mistake to
interpret it as a justification for the application of unproven wind turbine design to the
harsh offshore conditions.
The success of a design process is largely dominated by the organisation of the design
team as the process and organisation can hardly be separated. Figure 3.10 proposes
a scheme for the goal and structural coordination of the integrated design approach.
The management is done by a ‘team of cluster leaders’ in which every partner is
represented and all major decisions which affect the design of the system are taken by
the team. Each person in this group manages a design team which is responsible for
the development of a particular sub-system. It should be attempted to reduce the
number of relations between the various sub-systems by suitable selection of the subsystems. This will reduce potential design conflicts and will increase the flexibility of a
design team. Decisions made by the ‘head design team’ should be carefully and
consistently reported as they serve as ‘requirements’ for the sub-system designs of all
teams involved.
Figure 3.10 illustrates how the sub-system design is controlled with respect to the
project goals by means of five aspect systems: levelised production cost, adaptation on
site conditions, dynamics and structural reliability consideration, installation and
commissioning effort and OWECS availability. The aspect systems are resolved into
qualified and controllable criteria. Procedures are available to evaluate the design. The
engineers in the different disciplines involved need assistance and tools for judging
intermediate results during the design process. Examples include an overall cost model
[3.14], a simulation tool for the operation and maintenance behaviour [3.15] and codes
for design calculation of the entire OWEC (Chapters 6 and 8).
The complex system is split-up into sub-systems, each setting its design requirements.
These requirements are then input for each design team to provide a satisfactory
solution. If there are problems in the generation of such a solution the requirements
should be re-evaluated by the ‘head design team’ as the changes will influence the
design of all sub-systems (but probably not to the same extent).
Section 3.6 and Figure 3.11 summarize the first application of the integrated design
approach during the design of a 300 MW offshore wind farm within the Opti-OWECS
study. Probably the approach will also be considered in some proposed offshore
projects in the Netherlands including the 100 MW Nearshore project, the Afsluitdijk
project and the development project DOWEC [3.16]. Consortia covering the expertise
of the different
Elements
Subsystems
Aspect
systems
Goals
System
OWECS
installation
Wind farm
layout
Switch gear,
transformer,
etc.
Cable
OWECS
availability
Wind farm
operation
Maintenance
design
System
reliability
Operation & maintenance
(infrastructure & strategy)
Installation &
Commissioning
effort
Lifetime
Figure 3.10: Controlling of the integrated design approach
Re-design
wind turbine
Support
structure
Grid connection
and farm layout
Dynamics &
Structural reliability
Support structure
and installation
Adaptation to
site
Re-design
rotor blades
Wind turbine
Levelised
production
costs per kWh
Profit
Power quality,
quantity & persistence
Delivery of substantial electricity
Engineers
Cluster
leaders
Project manager
& cluster leaders
& expert staff
Client
Political
order
39
40
Chapter 3
technical and financial fields are developing entire offshore wind farms or an offshore
wind energy converter in these projects. Elements of the integrated design approach
were also applied during the realisation of the Utgrunden offshore wind farm in the year
2000 (Section 9.1) [3.17].
Within the integrated design approach specific offshore wind energy standards, [3.18],
are used and certification is done with respect to the particular site conditions.
The following sub-sections evolve four complementary variants of the integrated design
approach each emphasising particular aspects including economics, operation and
maintenance aspects, overall dynamics and radical design solutions. For the further
elaboration of this thesis the part dealing with dynamics is most important. In reality,
features of different variants will be combined according to the actual requirements.
3.4.2
Economic optimisation
The best economic performance of an offshore wind farm is achieved by a balance
between capital cost, operation and maintenance (O&M) costs, availability and
associated energy yield.
Important design parameters are interacting with certain sub-systems and their
performance in different ways. For instance, an increased hub height results in a higher
energy yield but also in higher support structure cost, installation and maintenance
costs. More exposed sites offer more energy and potentially lower energy costs but
Integrated design approach
Optimised maintenance strategy
- considering entire wind farm,
- wind energy and offshore technology
- selection of concept and site
by overall criteria (e.g. cost model)
- consideration of site conditions,
- adapted to wind turbine reliability,
- use of modified jack-up platform
Wind turbine
- robust 3 MW concept,
- reduced failure rates,
- improved maintainability,
- high availability
Costs savings of wind turbine
and support structure
- increased rotor speed,
- low turbulence,
- low hub height (dynamics,
installation, maintenance)
- transformer in nacelle
Monopile support structure
- viable for demanding North Sea
(up to 20 m depth)
- innovative pile - tower joint,
- correction of inclination,
- lower fatigue due to tapered pile
Advanced dynamic analysis
of entire system
- soft-soft support structure,
- compromise between wind
and wave response,
- lower weight and costs
Figure 3.11: Innovative features of the design solution for a
300 MW offshore wind farm
41
have to be paid for by larger investments, higher O&M cost, lower availability and
greater transmission losses.
In an integrated design approach such relations are considered from the outset. The
design effort can be directed towards crucial cost elements, e.g. operation and
maintenance costs. During the course of the design process the sophistication of the
cost evaluation should increase from simple estimates of the cost-breakdown and
levelised production cost in the feasibility study, to an engineering cost model for the
entire OWECS [3.14] during the conceptual design and the detailed evaluation in the
final design phase. After each stage, solutions for an entire offshore wind farm at a
particular site should be evaluated.
3.4.3
Design for reliability, availability, maintainability and serviceability
Challenging design targets for reliability, availability, maintainability and serviceability
(RAMS) [3.19] of an optimised offshore wind farm can only be met by an integrated
design approach.
The targets with respect to the price, quantity, quality and persistence of the delivered
electrical power do not only guide the technical design but also the design for reliability
and availability, i.e. failure states and rates, and the design for maintainability and
serviceability, i.e. rate, ease and cost of repair and regular service. Since there is only
a limited possibility to use experience for the development of an integrated maintenance
approach for OWECS, a knowledge-based methodology must be applied.
Likewise for the structural design, the following specific targets are set and are
effectively controlled in all phases of the design process.
• reliability target, i.e. the probability that the system is able to fulfil its functional design
targets,
• availability target, i.e. the fraction of time that the system is able to operate as
intended under given external conditions,
• maintainability target, i.e. the probability that a malfunctioning system can be brought
back into operation within a given time,
• serviceability target, i.e. the ease and costs at which regular (scheduled) service can
be carried out and specification of the fraction of time and the costs needed for
service.
The principal relation and development of design targets and design specification during
the different design phases are illustrated in Figure 3.12.
The specification of the functional design targets, the RAMS targets, and the design of
the installation and maintenance concepts should first be addressed in an integrated
way on the systems level during the feasibility study. Only then the best solution will be
achieved. The targets on the systems level should then be translated in a consistent
way into specifications on the sub-system level during the conceptual design. Next for
the detailed design the specifications have to be determined also on the component
levels. Finally the sub-components are treated in the same manner during the design
specification phase.
After the assessment of the local consequences of the specifications on a certain level
of the system, by a simplified calculation of reliability and availability, the results should
be evaluated on the system level. Such a process of continuous evaluation and reevaluation of the targets and specifications on all the design levels assures the optimal
design of the OWECS.
42
Chapter 3
RAMS targets
OWECS level
OWECS evaluation
with
RAMS design tools
Simplified R&A
calculation
System level
Design specification
OWECS level
Feasibility
study
Sub-system
level
RAMS targets
sub-system level
Local R&A
calculation
sub-system level
Conceptual
design
RAMS targets
component level
Component
level
Local R&A
calculation
component level
Detail
design
Sub-component
level
RAMS targets
sub-component level
Local R&A
calculation
sub-component level
System
specification
Figure 3.12: Relation between design targets, specifications and evaluation in the
design process for reliability, availability, maintainability and serviceability (RAMS)
(R&A = reliability and availability)
With respect to the further assessment of the possibilities to achieve the RAMS targets
it is worthwhile to apply design tools for reliability and availability calculations. Such
design tools range from simple probabilistic techniques for simplified reliability and
availability calculations based upon generic failure rates (small, inner iteration loops in
Figure 3.12), to more advanced Markov chain modelling. For complex systems such as
an offshore wind farm the most often applied technique for evaluation is Monte-Carlo
simulation (large, outer iteration loops in Figure 3.12).
It is evident that design solutions of the OWEC unit design level will have
consequences for the design of the OWECS maintenance concept and vice versa. But
an integrated approach of the design of the installation process and the maintenance
process will be beneficial for achieving the best solution. An example is the possibility
of using a general purpose or purpose built/modified installation infrastructure, such as
platforms, cranes and boats afterwards as maintenance infrastructure.
43
3.4.4
Overall dynamics
System dynamics and dynamic interactions between sub-systems are important and
should be considered in an integrated manner during the different design phases.
Examples of the treatment of dynamics of offshore wind energy converters (OWEC)
include:
Feasibility study
• (qualitative) compatibility of support structure and wind turbine concepts
• (qualitative) compatibility of support structure concepts and sites,
• assessment of the order of magnitude of aerodynamic and hydrodynamic loading for
extreme as well as fatigue conditions,
Conceptual design
• sensitivity analysis of dynamics with respect to soil properties,
• parameter studies on dynamic loading in frequency (or time) domain,
• simultaneous optimisation of wind turbine concepts (e.g. rotor speed and diameter,
blade and drive train layout) and support structure concepts (e.g. stiffness and hub
height),
Detailed design
• detailed dynamic analysis of OWEC with design tools in the time domain (especially
if fatigue is governing),
• fine tuning of dynamics,
• investigation of effects due to the variability of site parameters within the wind farm.
In Chapter 10 ‘Design optimisation by means of tailored dynamics’ demonstrates the
consideration of dynamics during the design process of an offshore wind farm.
3.4.5
Radical design for offshore requirements
The use of unconventional designs might provide a major benefit for the entire OWECS
under the condition that such radical designs are governed entirely by the offshore
requirements rather than by adapting onshore experience for the offshore situation. For
instance, the wind turbine would be designed consistently for offshore application from
‘scratch’ rather than ‘marinising’ an onshore design.
Without an assessment of their viability, some examples are listed here for illustration
purposes only.
The Multibrid concept of the engineering consultant aerodyn Energiesysteme GmbH
(Figure 3.13) introduces a novel wind turbine concept and several innovative features
in order to meet the design requirements of large offshore wind energy converters. In
contrast, most wind turbine manufacturers are following a less risky approach by upscaling of a certain design concept particular to the company. The integrated drive train
with a single stage gear and a medium speed generator of the Mutlibrid design
facilitates reduction of failure rates by minimising the number of components, avoidance
of high speed gears and bearings and complete capsulation of the entire nacelle. The
compact design, innovations in the blade technology and yaw system lead to an
extremely low tower head mass in the 5 MW power class [3.20].
Unconventional support structures are proposed by the multi turbine concept [3.21] and
the Multi Unit Floating Offshore Wind Farm (MUFOW) concept (Figure 2.4) [3.22].
44
Chapter 3
Water cooling of all
components
Single-stage
planetary
gearing
Only one
sealing
ring
towards
the
outside
Slowly rotating
generator
Integrated
machine set
Planet carrier
with
integrated
rotor bearing
Internal yaw bearing
Geometrical arrangement
Seawater heat-exchanger at foot of tower
Figure 3.13: Radical wind turbine design of the Multibrid concept
Ultimately, such a radical approach might be the most promising. However, unproven
or just ‘big’ designs, do not lend themselves well for application in the demanding
offshore environment and a different lesson can be taken from onshore development.
The large success of wind energy technology during the last decade was achieved only
by an evolutionary technical development. It required several years until the two major
innovations, i.e. variable speed - variable pitch control and direct drive technology,
showed commercial success. None of the radical designs of the 1980s and early 1990s,
including prototypes of multi-megawatt turbines, single bladed machines, Vertical Axis
Wind Turbines (VAWT), proved to be viable.
In summary, the radical approach is hardly considered feasible at the moment but may
be required for very large offshore wind farms with a capacity in the gigawatt range.
3.5
Implementation
The previous section described the philosophy behind and the organisation of the
integrated design approach. In the following one possible implementation during the
different stages of the design process is presented.
The procedure is based upon the experience gained during the Opti-OWECS study.
Emphasis is given to the sequence and interaction of the different parts of the work
contents rather than to organisation, scheduling and resource management. Only the
relations between the technical steps are considered here, no interaction with the
‘customer’ or the ‘public’ is taken into account. However, issues as public acceptance
and permission can have a significant and sometimes unexpected impact.
45
Start of design process
Step 0: Project identification
Step 0:
Identification
Objectives
Problem statement
Step 1.1 a):
Pre-selection of initial sites
Step 1.1 b):
Pre-selection of initial
sub-system concepts
Step 1.2:
Evaluation of OWECS concepts
and sites
Step 1:
Feasibility
study
Step 1.3:
Selection of (=>2) concepts
and (>2) sites
Function
Overall realization plan
Step 2.1 a): Conceptual sub-system design
for all OWECS concepts and sites
Wind
turbine
Support
structure,
installation
Grid
Operation &
connection,
maintenance
farm layout
Step 2.1 b):
Evaluation
OWECS tools
Step 2.2:
Evaluation of concepts and sites
Step 2:
Conceptual
design
Step 2.3:
Selection of final concept
and site
Preliminary Design
Concept description
Step 3:
Detail design
Step 3: Detail design
Final Design
Step 4:
Specification
Step 4:
Elaboration / Specification
System Specifications
End of design process
Figure 3.14: Example design process of the integrated design approach
46
Chapter 3
Wind turbine
Farm layout
Support structure
Grid connection
Installation
procedure
Operation &
Maintenance
major interaction
minor interaction
Figure 3.15: Interactions between sub-systems during the design process
The process (Figure 3.14) starts with the project identification (Step 0) and the
statement of the project objectives.
In the first step a feasibility study is performed which results in a selection of a number
of concepts related to certain sites (Step 1.3) and the overall realisation plan. The
choices are based on the separate pre-selecting of sites and sub-system concepts (Step
1.1) and the evaluation of the combination of sites and OWECS concepts (Step 1.2).
In the conceptual design (Step 2) the chosen combination of sites and OWECS
concepts are further developed on a sub-system level (Step 2.1a). In parallel tools for
the evaluation of the OWECS performance and the control of the aspect systems are
developed or updated (Step 2.1b)
Again an evaluation (Step 2.2) and selection of the final concept at the final site (Step
2.3) takes place. Once the outcome of this phase is given, the client does not have
opportunities to change concepts or scope of work without incurring substantial delay
and extra costs. Therefore the outer right loop from the conceptual design to the
problem identification is indicated only as a dashed line. During the detailed design in
Step 3 the ‘dimensions’ of the design are fixed and the relation between the subsystems is generated.
System specifications are worked out in the final, fourth step.
While during the conceptual design only major interactions between sub-systems are
considered a more demanding treatment is required during the structural design phase.
Now the interactions between the sub-systems are taken into account in a form as
complete and practical as possible (Figure 3.15) and the solution is governed by
quantified, overall criteria.
3.6
47
First application
The first application of the integrated design approach during the Opti-OWECS is
summarised to demonstrate the feasibility and the prospects for future applications.
Here only the procedure and the main experiences rather than details of the actual
solution [3.23] are presented.
An offshore wind farm was developed against the background of three objectives.
Firstly, the improved understanding of the principles underlying the design of OWECS
gained during the course of the project was to be demonstrated by a practical solution.
Application of promising innovations for large-scale utilisation, e.g. novel installation
methods, consideration of operation and maintenance aspects, integrated design
approach, was for this research project more important than achieving the absolute
economic optimum. Secondly, during the design process, areas of poor understanding
were to be identified and appropriate solutions were to be developed. Finally, the
economic feasibility of large OWECS was to be investigated.
Project identification
The initial phase of the project identification comprised three aspects:
• establishment of the project group: an international cooperation of industrial
engineers and researchers from the wind energy field, offshore technology and
power management
• determination of project conditions (i.e. objectives and work programme)
• formulation of the particular design conditions, e.g. wind turbine size of 3 MW or
larger, neglect of the details of the onshore grid connection.
Feasibility study
During the feasibility study, a broad inventory of all relevant aspects and concepts was
made and pre-selections for the conceptual design were identified. Furthermore, a
terminology appropriate to OWECS was established in order to promote smooth
communications (Appendix A).
The identification of seven distinct reference sites in northern European waters was
carried out in parallel with further investigation of the sub-system concepts and the other
essential features of overall dynamics and operation and maintenance. The water depth
of the considered sites ranged from 8 m to 25 m. The majority of the sites can be
characterised as genuinely offshore rather than inshore or nearshore sites, as is the
case for existing offshore wind farms.
Based upon a qualitative OWECS evaluation the following sub-system concepts were
selected for further development:
• The 3 MW 80 m diameter Näsudden development line was chosen as reference
turbine since it is the only one in the multi-megawatt league with a reasonable
operational track record. In addition, the turbine rating of about 3 MW extended the
state of the art of other offshore projects under development during the end of
1990s. Another justification for this choice was the availability of the entire set of
design data and the industrial interests within the project consortium.
48
•
•
•
•
Chapter 3
two wind turbine concepts (geared - fixed speed, direct-drive - variable speed),
rotor variants with diameters between 80 and 100 m and different rotor speeds,
distinctly different combinations of support structure configuration and installation
procedure, dynamic characteristics (stiff-stiff, soft-stiff and soft-soft 3) and site (North
Sea and Baltic Sea),
standard options for grid connection and wind farm layout.
Conceptual design
The conceptual design phase was carried out mainly in parallel with work on subsystems development and with the extension of the OWECS tools for cost modeling,
simulation of the operation and maintenance behaviour, structural reliability
considerations and overall dynamics.
Improved knowledge of particular OWEC aspects gained during this phase led to the
consideration of a monopile support structure in addition to the gravity monotower and
the gravity lattice tower (lower row in Figure 3.16). The extreme design wave together
with a relatively low water depth resulted in a very significant heave force on the gravity
monotower and required a large amount of ballast which is costly. Furthermore, both the
soft-stiff monotower and the stiff-stiff lattice tower suffered considerable aerodynamic
fatigue loads owing to their fundamental eigenfrequency being close to the blade
PILED - BRACED MONOTOWER
Installation: Lifted
GBS - LATTICE TOWER
Installation: Floated Out
GBS - BRACED MONOTOWER
GBS - MONO TOWER
MONOPILE
Installation: Lifted or
Floated Out
Figure 3.16: Examples of support structure concepts
(GBS = Gravity-base System)
3
Three design solutions exist for wind turbine towers depending on the ratio between the fundamental
eigenfrequency fO and either the rotor frequency fR or the blade passing frequency f b = N b @ f R : soft-soft,
i.e. fO < fR , soft-stiff, i.e. fR < fO < fb and stiff-stiff, i.e. fb < fO . In practice, stiff-stiff designs are no
longer built for the current generation of wind turbines.
49
passing excitation frequency and the constant speed - variable pitch concept. In
particular, stress concentration at the tubular joints and the stiff characteristics of the
lattice design, chosen well beyond the range of dynamic wave excitation at the exposed
North Sea site, showed conflicts with the dynamic wind turbine loading. The dynamics
of the three concepts are further discussed in Chapter 9.
Several results were directly related to the integrated approach:
• integrated development of support structure concepts and installation procedure
• simultaneous optimisation of wind turbine (e.g. rotor speed, blade layout) and
support structure (e.g. stiffness) with the main goal of reduction of aerodynamic
fatigue loads (Section 10.2.2)
• consideration of overall dynamics in the support structure design
• development of optimal O&M strategies based on Monte-Carlo simulations,
• development of structural reliability analysis for the support structure
Next, the cost model, developed elsewhere during the project, was used to evaluate
different offshore wind farms assembled from the developed sub-system concepts at
the seven pre-selected sites from the feasibility study. For the same OWECS concept
the energy costs differed up to 60% between the seven compared sites, which initially
were all considered as promising. Furthermore, a reduction of the energy cost of about
20 % was found by application of a support structure of the monopile variety instead of
a gravity monotower type. The economics together with some other criteria led to the
selection of the final concept and the related site.
Detailed design
During the detailed design phase, the selected concept was developed further and
interactions between the sub-systems were fully considered.
This integration facilitated several innovations, which represent significant steps towards
mature offshore wind energy technology (Figure 3.11):
• An appreciably high offshore farm availability of 96.5% was achieved through a
rational approach comprising a gradual improvement of the turbine’s reliability and
maintainability with respect to the current onshore state of the art and an innovative
operation and maintenance solution. The latter included both an optimised operation
and maintenance strategy and a cost-efficient solution to the ‘cranage problem’ in
the form of a permanently and quickly available self-propelled modified jack-up.
• Close cooperation between structural design and dynamic simulations of the OWEC
facilitated a soft-soft monopile design even for 20 m water depth (LAT) in a
demanding Southern North Sea environment with firm soil conditions. Balancing the
combined aerodynamic and hydrodynamic fatigue loads, reduction of the waterpiercing cross section and consideration of weight and installation issues were the
main reasons for this choice. This is remarkable because the fundamental
eigenfrequency of 0.29 Hz lies well below the rotor frequency of 0.37 Hz but just in
the region of significant dynamic wave excitation extending between 0.04 and 0.5
to 1 Hz.
• Consideration of tower and foundation as one system resulted in a more appropriate
and economic design than the separated treatment of tower and foundation.
• Significant cost reductions were possible by innovative installation methods for
gravity systems which were floated and towed as entire OWEC unit to their final
50
•
•
Chapter 3
destination. Conceptual solutions were developed for both floating and lifting
installation of monopile structures. The variety of offshore equipment available,
including jack-ups, crane vessels, barges, etc. opens several opportunities which
have to be judged case by case. Also, open questions remain, for instance on the
sea fastening for the vertical transportation of the tower and wind turbine unit.
The aerodynamic efficiency of the wind farm, the cable costs of the grid connection
and the space requirement of the OWECS were well balanced.
Placement of the OWEC transformer in the nacelle was chosen after consideration
of wind turbine, support structure, grid connection and maintenance aspects. Most
important for this choice was that expensive modifications of either the tower
diameter or of the access platform are avoided and that the dynamic characteristics
remain nearly unchanged.
It is worth noting that neither during the detailed design phase, nor after the final
evaluation of the design solution, major revisions of the design were required. The main
reason for this was that the conceptual design had already been carefully examined with
respect to technical feasibility and economic performance.
Beside the technical feasibility and economic prospects the Opti-OWECS study has also
been demonstrated that such a methodology can be managed within an international
cooperation of scientists and industry engineers from wind energy technology, offshore
technology and power management.
51
Description of the Offshore Environment
CHAPTER 4
DESCRIPTION OF THE OFFSHORE ENVIRONMENT
Various natural and man-made environmental conditions are affecting offshore wind
farms (Figure 4.1). The most relevant for our purposes are the meteorologicaloceanographic conditions, the so-called ‘met-ocean’ parameters, and the soil conditions.
This chapter describes the environmental parameters while the next chapter deals with
the modelling of offshore wind energy converters and the environmental impact.
Only an overview can be provided here. Therefore the extensive usage of formulas has
been avoided. For a more comprehensive discussion the interested reader may consult
the references and particularly the design guidelines of the UK Department of Energy
[4.1] and textbooks on meteorological boundary layer theory [4.2].
The different met-ocean parameters are discussed in Section 4.1 where special
attention is given to wind and waves. The different needs of offshore wind energy
applications compared to land-based wind energy converters and offshore platforms are
highlighted. Section 4.2 is devoted to the correlation of wind and wave conditions, while
Section 4.3 concerns the soil conditions and some auxiliary environmental effects such
as corrosion, ship impact and influences arising from the energy consumer.
ambient turbulence
grid interaction
salinity, humidity
& temperature
wake turbulence
(breaking) waves
currents
tides
earthquake
lightning
operational
& accidental loads
icing
ship impact
(breaking) ice
marine growth
scour
foundation
behaviour
Figure 4.1: Environmental impacts on an offshore wind farm
52
Chapter 4
Table 4.1: Met-ocean parameters relevant for the design of OWEC
Wind
• directional long-term distribution of mean wind speed at hub height
• extreme wind speeds (averaged over 10 min, 1 min, 3 s; RP1, RP50)
• vertical profile (roughness length or wind shear exponent)
• distribution of stability conditions
• ambient turbulence intensity at 15 m/s mean wind speed
• turbulence spectrum, coherence function and integral length scales
Waves
• extreme wave crest height
• extreme wave height, direction and range of associated periods (RP50)
• extreme sea state parameters (RP50)
• (directional) scatter diagram of significant wave height and zero crossing or
peak period
• wave spectra and directional spreading functions
Wind and wave correlation
• long-term correlation of wave parameters and mean wind speed
• distribution of weather windows
Currents
• extreme current speed and direction
• vertical profile
Water depth and sea level variations
• water depth below lowest astronomical tide (LAT)
• tidal range
• extreme still water level variation, e.g. surge
Icing and sea ice
• design thickness of snow and ice on the structure
• sea ice design parameters (if relevant)
Temperatures
• extreme air temperatures
• extreme sea temperatures
Marine growth
• type and density of growth
• annual growth rate and terminal thickness profile
RP1 = annual return period, RP50 = return period 50 years
53
4.1 Met-ocean parameters
Harsh weather conditions, hindered access for inspection and unmanned operation for
long periods, along with the need for cost-efficient design solutions result in great
importance of the description of the environment (Table 4.1).
The specific properties of offshore wind farms do not allow unproven application of the
different experiences from both the offshore and the wind energy technology.
For instance, offshore wind farms are generally located in more shallow waters and
closer to shore than most offshore petroleum installations resulting in different load
characteristics and more complex hydrodynamic behaviour. Wind loads and
hydrodynamic loads are often of the same order of magnitude. Generally fatigue is more
important. The highest loads do not necessarily occur during 50 year extreme storm
conditions since extreme aerodynamic loads during production could be more
demanding.
Onshore wind turbines are designed according to a few standardized wind classes
appropriate for a variety of sites. The offshore wind climate, however, differs
considerably from land and between locations. Hence, for large projects a site specific
offshore design is needed from a technical and economical point of view.
Wind
4.1.1
Wind is continuously changing with time by nature. Figure 4.2 illustrates the frequency
and energy contents of the different types of wind phenomena. A spectral gap opens
between the macro-meteorological effects due to the gradient between pressure
systems and semi-diurnal winds on the low frequency side and the micro-meteorological
variations owing to thermal and roughness induced turbulence on the high frequency
side. The low energy content between periods of some hours and about five minutes
enables the separation of long-term and short-term statistics.
frequency [Hz]
0.01
0.1
log(PSD* frequency) [m2/s2]
pressure
systems
turbulence
semidiurnal
120
24 12
macro-meteorological
range
1
spectral
gap
0.1
period [h]
micro-meteorological
range
Figure 4.2: Qualitative wind energy spectrum [4.3]
54
Chapter 4
In comparison to their counterparts on land, offshore winds are generally characterised
by higher annual average wind speed, higher extreme mean wind speeds but lower
turbulence and gust factors along with a smoother wind shear and more stable wind
directionality.
For most sites and with a reasonable turbine spacing, this results in lower aerodynamic
loads and offers opportunities for design optimisation. At exposed sites the benign
effects are offset by both the more frequent high wind speeds, which dominate the
fatigue, and stronger extreme gusts governing limit state. Section 9.2 quantifies site
specific loading in more detail.
The following discussion focuses on wind aspects relevant for load calculations. Other
specific meteorological effects are of importance for reliable predictions of the wind
resources and for extrapolating measured wind speeds to larger heights. References
[4.4 - 4.8] provide further information on daily and seasonal variations, the effect of the
internal boundary layer thickness, atmospheric stability corrections, vertical wind
profiles, wind speed dependent roughness, ambient turbulence and low level jets.
Wind speed distribution and wind profile
Figure 4.3 compares the wind speed distribution between two fictitious but typical sites
onshore and offshore. At sea the annual average wind speed, Vave , is greater and the
shape of the probability density function of the mean wind speed is shifted to higher wind
speeds corresponding to a larger Weibull shape factor, k. Just for comparison, the
Rayleigh distributions with k = 2 are plotted as solid lines without markers for both sites.
Generally on sea the annual average wind speed and the shape factor are larger (see
solid lines with markers). Both properties results in a larger probability of high wind
speeds and favour a high specific rating for OWEC in terms of rated power per unit
swept area. This can be realised for instance by a greater generator capacity at the
same rotor diameter compared to onshore siting.
The reduced surface roughness at sea results in a steeper vertical wind profile and a
lower turbulence intensity. Small wavelets or ripples on the water surface rather than the
wave height itself determines the roughness. Charnock proposed a constant ratio
between skin friction velocity and roughness, valid only for the equilibrium state between
wind and wind generated waves [4.9]. Often short fetch, i.e. the distance the wind is
acting over the sea, or short duration of the wind condition result in circumstances which
are not fully developed, in which the Charnock constant is no longer applicable.
However, in most cases a roughness length of 0.2 mm [4.10] is a reasonable
assumption compared to 0.01 - 0.25 m for sites on land [4.11] 1.
For engineering purposes the wind speed profile is expressed by a logarithmic relation
between height over ground and roughness, z0 , [4.11, 4.12] or by a power law as
function of non-dimensional height with the wind shear exponent, α, as a parameter
[4.13 - 4.15]. Both relations are valid only for neutral atmospheric stability conditions
1
The Danish wind turbine standard DS 472 recommends z0 = 0.001 under normal and 0.004 m under
extreme offshore wind conditions [4.12]
55
0.06
Onshore: V ave = 7 m/s, k = 1.75 (Weibull)
Onshore: V ave = 7 m/s
k = 2 (Rayleigh)
probability [-]
0.05
0.04
Offshore: V ave = 8.5 m/s
k = 2.2 (Weibull)
0.03
0.02
Offshore: V ave = 8.5 m/s
k = 2 (Rayleigh)
0.01
0.00
0
2
4
6
8
10
12
14
16
18
20
22
24
wind speed at hub height [m/s]
Figure 4.3: Wind speed distributions for typical sites
onshore (Vave = 7 m/s) and offshore (Vave = 8.5 m/s)
(solid: Rayleigh distribution, solid with markers: Weibull distribution)
120
Onshore:
V ave = 7 m/s at 80 m
z 0 = 50 mm
100
V ave = 8.5 m/s
at 60 m
z 0 = 0.2 mm
80
height [m]
Offshore:
60
40
20
0
5.0
6.0
7.0
8.0
9.0
annual average wind speed [m/s]
Figure 4.4: Logarithmic wind profiles for typical onshore and offshore sites
(z0 = roughness length)
26
56
Chapter 4
stability conditions when thermally induced turbulence is negligible compared to
mechanically induced turbulence. Neutral conditions are less frequent offshore but they
are still dominating in the range of high wind speeds which are most relevant for the
mechanical loads. Figure 4.4 illustrates how the wind profile develops at two typical
sites. Offshore the hub height is commonly chosen lower (Table 2.1) since relatively little
extra energy can be gained by a taller structure, which is relatively more expensive than
on land.
Atmospheric turbulence and wind field simulation
Turbulent gusts are caused by eddies in the wind flow. For the wind field simulation it
is assumed that they travel with the mean wind speed according to the commonly
applied frozen turbulence hypothesis (Figure 4.5). The eddies have different
frequencies, amplitudes and phases and may intersect each other. The characteristic
lengths of the turbulence structures, denoted as integral length scales, are of the same
magnitude as the rotor diameter of large wind turbines. The rotational period is
commonly smaller than the time for a turbulence bubble to pass the swept area.
Consequently most of the gusts affect only parts of the rotor but are seen several times
by blades. This ‘rotational sampling’ or ‘eddy slicing’ transforms the wind spectrum seen
by a rotating observer leading to a lumping of excitation energy at the rotor frequency
1P and its multiples (Figure 4.6).
The temporal and spatial turbulence properties can be represented by power spectra
associated with the three velocity components and a coherence function accounting for
the spatial correlation. Based upon the work of Shinozuka [4.16] and Veers [4.17], the
auto and cross spectral density functions are determined at discrete points of the swept
area [4.18] or on the rotor blades [4.19] from which random representations in the time
domain are generated by Inverse Fast Fourier Transform (IFFT).
In this thesis the longitudinal turbulence is described by a von-Karman spectrum and an
exponential coherence function with aid of the SWING 3 code [4.20]. Recently it has
become common to fully account for all three components of the turbulent wind field
[4.15]. Here only the longitudinal component is simulated stochastically. The effect of
the other components is modelled in a simplified manner by a sinusoidal variation of the
yaw misalignment of ± 300 during the duration of the dynamic simulations [4.21].
Ambient and wake turbulence
The lower roughness at sea reduces the atmospheric turbulence intensity, i.e. the ratio
of the standard variation of the instantaneous wind speed and the mean wind speed,
and stretches the integral length scales of turbulence. Measured values of the
turbulence intensity, range around 7 - 9% at 15 m/s for neutral conditions offshore [4.22,
4.23]. Larson and Jørgensen proposed an empirical relation for the design turbulence
intensity based upon evaluation of two Danish shallow water sites and the fatigue
behaviour of different materials [4.24]. At 15 m/s mean wind speed this results in a figure
of 9.7%; however, above 20 m/s unrealistic high values are obtained.
Generally, the same power spectra and coherence functions are considered onshore
and offshore but with different design turbulence intensities. Figure 4.7 compares the
design turbulence intensity as function of mean wind speed at the two sites onshore and
offshore. At low wind speed thermal turbulence and non-neutral conditions are prevailing
and higher turbulence is observed. For higher wind speeds, the mechanical turbulence
owing to roughness dominates and the turbulence intensity decreases.
57
The dynamic loading in a wind farm increases significantly due to the wake of upstream
turbines. Especially when the swept area of a turbine is partly covered by a wake, high
load fluctuations are observed. Measurements at the Vindeby offshore wind farm
indicate that the relative increase in wake turbulence is considerably higher offshore and
is more persistent downstream due to the lower ambient turbulence. The observations
showed a machine to be affected only by the directly adjacent turbines rather than by
all upstream turbines [4.25]
For engineering purposes it is convenient to account for the different effects of the wake
on the wind turbine components by an increase of the ambient turbulence intensity
which results in approximately the same fatigue damage as in the actual operation at the
most unpleasant position within the wind farm.
Frandsen et al. [4.25] introduced a relatively simple correction which considers the wind
turbine thrust coefficient, mean wind speed and spacing ratio, only. This model is
evaluated for the pitch-regulated wind turbine used within the Opti-OWECS study in
Figure 4.7. The wake effect is relatively small offshore if a large spacing of 8 to 10
diameters is applied.
Other research pointed out an effect of the slope of the S-N curve of the material under
investigation [4.26, 4.27]. For instance, rotor blades made from glass fibre reinforced
plastic are more prone to high cycle fatigue and are consequently more sensitive to
wake effects than steel components, e.g. the tower.
Recent investigation suggest that the wind turbine of large wind farms are acting as
roughness obstacles and are inducing a background turbulence that at least parly
compensate low ambient turbulence [4.27].
Extreme wind gusts and directional changes
Extreme wind conditions relevant for design are distinguished by the operational mode
of the wind turbine to be either parking, i.e. standing still, idling or failure state, or power
production.
The governing load for parking results from extreme storm conditions with a 50 year
return period. Here the traditional approach of a quasi-static calculation of the extreme
dynamic pressure corrected with a dynamic amplification factor [4.28, 4.29], is now
replaced by more accurate and less conservative time domain simulations of the
response under a turbulent wind field.
The practice with respect to extreme wind conditions during production is in contrast to
the sophisticated description of the stochastic behaviour under fatigue conditions. All
wind turbine standards still assume simplified and deterministic situations, e.g. a given
variation of the mean wind speed by a gust with a ‘one-minus-cosine’ or ‘Mexican-hat’
shape. Such gusts are coherent over the swept area and combined with or without a
prescribed change of the wind direction. Other examples include an extreme vertical
and/or horizontal wind shear or extreme direction change [4.15]
Application of the above-mentioned standard simulation techniques for turbulent wind
requires unreasonable computations to find time series which includes an extreme gust
with a certain magnitude. Recently, a more adequate probabilistic description has been
demonstrated that embeds a prescribed gust amplitude in a background turbulence in
such a manner that the stochastic properties of the synthesised signal cannot be
distinguished from real turbulence [4.30]. The new approach applies methods which are
common in advanced offshore engineering [4.31] and is an example of a crossfertilisation between wind energy and offshore technology.
58
Chapter 4
z
y
x
Mean wind direction
Figure 4.5: Wind turbine rotor slicing though the temporal and
spatial structure of a turbulent wind field
Figure 4.6: Rotational sampling of gusts illustrated at the
transformation of the Kaimal turbulence spectrum from a
fixed to a rotating observer
59
30%
turbulence intensity [%]
Onshore
design turbulence IEC 61400-1 ed.2
I _15 = 18%, a = 2, accepted for > 5 D spacing
25%
20%
15%
spacing 10 D
spacing 5 D
10%
isolated turbine
Offshore
5%
ambient turbulence I _15 = 8%, a = 3
0%
5
10
15
20
mean wind speed at hub height [m/s]
25
Figure 4.7: Onshore and offshore turbulence intensities
(Functional relation of ambient turbulence acc. to IEC 61400-1 ed.2,
wake turbulence acc. to [4.25]: I_15 = turbulence intensity at 15 m/s
mean wind speed, a = slope parameter)
Table 4.2: Wind conditions according to Germanischer Lloyd [4.13, 4.14]
Wind condition
Onshore
Reference wind speed
(at hub height) Vref
Extreme gust Ve50
†
Offshore
identical type classes
Class III: 37.5 m/s
Class I: 50 m/s
Class II: 42.5 m/s Class IV: 30 m/s
1.4 @ Vref
1.2 @ Vref
Extreme operating gust
identical 13 m/s
Extreme change in wind
direction
(linearly interpolated)
V # 5 m/s : 180o
V = 15 m/s : 35o
V = Vref :
15o:
Partial safety factor
• environmental loads
• other loads
†
(F = 1.5
(F = 1.35
identical ((F = 1.1 - 1.5)
Germanischer Lloyd originally stated 5 s as averaging period while commonly 3 s
are used in accordance with IEC 61400-1 ed.2
60
Chapter 4
Owing to the lower turbulence offshore the Germanischer Lloyd has proposed less
severe gust factors offshore (Table 4.2). Nonetheless, the magnitude of the extreme
operational gust is maintained which is an important extreme load case for most pitch
regulated turbines.
4.1.2
Waves and currents
From the various hydrodynamic effects mentioned in the macro-scale energy spectrum
in Figure 4.8 the wind induced waves and tidal currents are the most important in
European coastal waters. Generally one has to distinguish between wind seas
generated by local winds and swells originating from waves travelling over a long
distance into the area of interest. The topography of the North Sea and the Baltic Sea
lead to lower significance of swell.
The period range of energy rich waves lies between 2 s and 20 s. Extreme waves have
long periods, typically 7 - 13 s for our purposes, since the energy of a sea state is
proportional to the fourth to fifth power of the dominant period and the square of the
wave height. Nonetheless, waves of moderate height are most important for the dynamic
wave loading since their frequency is closer to the eigenfrequencies of the structure and
their probability of occurrence is high.
The energy distribution of the wind generated waves corresponds to the macro-scale
wind spectrum in Figure 4.2. Low energy is observed between one and six hours. 2
Commonly a sea state is regarded stationary within a period of 3 hours. So, again shortterm and long-term statistics can be separated.
spectral density [m2.s]
frequency [Hz]
0.01
0.1
1
5
wind waves
tides
planetary
waves
swell
seiches
capillary
waves
surges and
tsunamis
672
24 12
1
0.1
period [h]
Figure 4.8: Qualitative energy spectrum of ocean waves [4.32]
2
The concentration of wave energy with periods of approximately 0.1 hour does not correspond directly
with the wind energy spectrum because seiches result from local resonance phenomenon of waves in
bays and harbours.
61
Wave spectra
The disturbance of the water surface elevation by random ocean waves is described by
semi-empirical energy spectra. In distinction to a turbulent wind field no coherence
function is required as the entire flow field is derived from a single point on the still water
surface only. The surface profile in the vicinity of the structure is composed from wave
trains and the kinematics of the water particles is determined entirely from the motion
of the water surface vertically above it. Various parametric forms of spectra for wind
generated waves can be found in the literature. A comprehensive overview is provided
by [4.33, 4.34].3
In order to understand some basic characteristics of wave energy spectra a simple wave
generation model is elaborated (Figure 4.9). Suppose an initially calm ocean and a
constant mean speed. The shear force and the differences of the dynamic pressure over
a crest and trough are generating perturbations of the free surface. Now waves with a
low phase velocity and corresponding high frequency and short wave length are excited
because of the large difference between the wind and wave phase velocity. The height
of a wave grows until it reaches the deep water breaking limit. Then kinetic energy is
transferred to other waves with lower frequency. The new waves are again enlarged by
the wind. This process of shifting energy down the frequency scale is continued until a
certain limit where wind speed and wave phase velocity are in equilibrium.
Therefore wave spectra (Figure 4.10) have a long trail towards the high frequency end
but a steep drop in energy below which wave energy can only come from breaking high
frequent waves. The wind has to blow for a certain duration and waves need a
sufficiently long distance of propagation to generate a fully developed sea. For high wind
speeds, both conditions are relatively rare, hence most of the related sea states are
either duration limited or fetch limited.
Assuming a Gaussian distribution of the water surface elevation, the random waves can
be resolved in a superposition of an infinite number of regular wave trains with various
frequencies and wave lengths. Each elementary wave train is of sinusoidal form with an
infinitesimally small amplitude. In most cases it is assumed that all waves are
propagating in the same direction. This is denoted as long-crestedness owing to the
uniform lateral wave field. In reality waves from different directions are interfering and
any sea state is short-crested to a certain extent.
Wave spectra are separated in an omnidirectional frequency spectrum and a commonly
frequency-invariant function for the directional spreading if short-crestedness is
considered. The most relevant spectra for engineering purposes are the PiersonMoskowitz (PM) spectrum and the JONSWAP (Joint European North Sea Wave Project)
spectrum.
The Pierson-Moskowitz spectrum in its original form is based on one parameter only, i.e.
the hourly mean wind speed at 19.5 m height [4.36]. Assuming fully developed sea
3
In literature at least four different spectral periods are distinguished referring to the mean zero-up
crossing period, the spectral peak, the average period of the highest third of the waves and the mean
wave period. Particular care is required to identify them in a given formula for a certain spectrum.
Reference [4.33] provides accurate general definitions. For Pierson-Moskowitz spectra also conversion
formulas are given.
62
Chapter 4
states with infinite water depth, fetch and duration only, the spectrum is much too
conservative. Frequently a modified, two-parametric form [4.33] is considered for which
energy content and frequency range can be controlled by prescribed values of the
significant wave height Hs , i.e. the average of the 1/3 highest waves in a record, and the
zero up-crossing wave period Tz.
Fetch and duration limited sea states are better represented by the JONSWAP spectrum
[4.37, 4.38]. A third parameter, the peak enhancement factor γ, controls the amplification
of the spectral peak and governs the Pierson-Moskowitz spectrum for γ equal to one.
The peak enhancement ranges from one to seven with a mean of 3.3. Just for
comparison, Figure 4.10 shows different spectra for the same values of Hs and Tz. For
larger γ the energy is more concentrated and shifted to higher frequencies.
The choice of the wave spectrum depends also on the available set of sea state
parameters. In practice often only the scatter diagram for wave height and period exist
Figure 4.9: Sea state development [4.35]
90
Hs = 4.6 m
spectral density [m^2/Hz]
80
Tz = 8.4 s
JONSWAP (gamma = 7)
70
JONSWAP (gamma = 3.3)
60
Pierson-Moskowitz
= JONSWAP (gamma = 1)
50
40
30
20
10
0
0.050
0.075
0.100
0.125
0.150
cyclic frequency [Hz]
0.175
0.200
Figure 4.10: Comparison of Pierson-Moskowitz spectrum and JONSWAP wave
spectra for different peak enhancement γ but same value of Hs and Tz
63
and the Pierson-Moskowitz spectrum is used for fatigue calculations and the JONSWAP
spectrum with a mean value of the peak enhancement of γ = 3.3 is applied for extreme
conditions. This thesis emphasizes fatigue aspects and the Pierson-Moskowitz spectrum
is used unless stated otherwise.4
Directionality, directional spreading and wave misalignment of sea states
The sea state distribution included in a scatter diagram varies for different mean wave
directions owing to the associated differences of fetch, water depth and wind speed
distribution. In offshore engineering this so-called ‘directionality’ of sea states is either
neglected by averaging the wave characteristics of all directions, so-called
omnidirectional waves, or by analysing the response in 8 to 12 directional sectors.
Independent of the directionality the wind and wave direction are misaligned to a certain
extent and the sea states are showing a ‘directional spreading’ of energy or shortcrestedness. The various forms proposed for the directional distribution of wave energy
all consider a bell-shape spreading with respect to the dominant direction. One wellknown formula distributes the wave energy over ± 900 symmetrical to the main wave
direction by weighting the spectral density by a power of a cosine function. Without
further data a quite narrow distribution should be assumed which obtains 46% and 78%
of the total wave energy within a directional sector of ± 150 and ± 300, respectively [4.1].
In most circumstances it is conservative to neglect short-crestedness. However, during
production offshore wind energy converters experience aerodynamic damping of the
response in the direction of the rotor axis, which generally coincides well with both wind
and wave direction. At the same time only structural damping of the lateral response
exists. Consequently, the combined wind and wave response might be more onerous
if wave misalignment and/or directional spreading is accounted for (Section 9.6.1).
Extreme waves
In most cases establishment of extreme sea state conditions and associated extreme
design waves for, say, 50 year return period has to be done by extrapolation from a few
years of measurements. Such extreme value statistics are highly dependent on the
assumed shape of the low probability tails of the parent distributions, which are not well
known. The three parameter Weibull or the Fisher-Tippet I distribution are used with
reasonable success for extrapolation of extreme significant wave heights [4.38]. For
instance, Germanischer Lloyd [4.13] states two procedures to extrapolate the distribution
of wave heights and to establish an extreme design wave.
The standard treatment for common fixed offshore structures includes analyses of single
design waves with different associated wave periods and directions in conjunction with
a non-linear wave theory. Alternatively, the extreme sea state loading is investigated by
a stochastic approach.
4
Actually, this treatment is likely not conservative for typical OWEC support structures since the
physically more appropriate JONSWAP spectra would yield relatively more excitation energy at the
fundamental frequency of the structure than the used Pierson-Moskowitz spectra.
64
Chapter 4
Water kinematics and wave theories
The structural members of most offshore installations are small compared to the wave
length. Such structures are considered to be hydrodynamically transparent, i.e. the flow
field is not altered by the presence of the structure. Only when the diameter of a local
element exceeds 20% of the wave length the structure is called ‘hydrodynamically
compact’ and the flow is corrected for diffraction and refraction (Section 5.2.5). 5
Assuming a periodic or regular wave is helpful to understand some fundamentals of
irregular waves. More importantly, regular waves are considered in the analysis of the
extreme loading of offshore structures with low dynamic response.
A regular wave is approximately sinusoidal if the wave height, H, is small compared to
the water depth, d. The shape of the wave progresses with the celerity, i.e. wave length,
L, divided by wave period, T. The water particles normally move with a lower velocity on
elliptical orbits so that the travelling wave shape is caused partly by a horizontal and
partly by a vertical particle motion. In deep water the orbits are circular and the
maximum horizontal particle velocity at the water surface equals π·H/T while in more
shallow waters the orbits are more and more stretched in the direction of wave
propagation leading to considerably higher horizontal velocities than π·H/T. The diameter
of the orbits and consequently the associated velocities and wave forces are decreasing
exponentially with depth, d. At the sea bottom of shallow waters significant particle
velocities in the horizontal plane are observed (Figure 4.11).
Analytical wave theories are dealing only with a single, regular wave and are based
upon solutions of the potential flow problem and the associated boundary conditions.
Their validity is limited to a certain range of non-dimensional water depth, d/g·T2, and
non-dimensional wave steepness, H/g·T2 (Figure 4.12).
Linearisation of the boundary condition at the free surface is the base for the Airy or
linear wave theory [4.33]. It can be applied in a wide region, i.e. waves with small or
intermediate steepness in deep water and for waves with small steepness in
intermediate water. Linear theory is also the common base for modelling of irregular
waves by superposition of many elementary waves with small amplitudes.6 Free surface
effects and finite wave height are accounted for by a non-linear correction of the flow
field. Quite popular is the approach after Wheeler which considers the kinematics at the
instantaneous free surface to be identical to those originally calculated at the still water
level [4.39].
Stokes wave theories [4.33] are commonly used for extreme waves in deep water and
for intermediate depth. In shallow waters and for steep waves with a height up to about
90% of the breaking wave height the Stream function wave theory of different orders are
appropriate [4.40].
Within this thesis linear wave theory with Wheeler stretching is applied for sea state
analysis. Extreme regular waves are described by higher order Stream function theory.
5
For engineering purposes the consideration of diffraction effects is restricted to the use of the linear
wave theory.
6
Non-linear, irregular wave theories have been proposed but are difficult to apply and are not frequently
used for engineering purposes, yet.
Figure 4.11: Particle orbit of regular waves [4.38]
Figure 4.12: Ranges of suitability of regular wave theories [4.38]
( d = water depth, g = gravity acceleration, H = wave height,
L = wave length, T = wave period )
65
66
Chapter 4
α
Spilling breaker
. # 0.5
Plunging breaker
0.5 < . < 3
Surging breaker
. $3
Figure 4.13: Breaking wave types as function of non-dimensional
slope
ζ = tan α g T 2 2 π H
[4.41]
Shallow water effects and breaking waves
Due to the restricted water depth the generation of high waves is hindered or high waves
propagating from the open sea towards the shore become steeper. The wave height
increases while the wave period remains constant and the wave length decreases. The
more shallow the water the more non-linear such waves behave. The crest height
increases while the through decreases and water particle velocities are significantly
higher in the wave top than in the through (Figure 5.15).
Compared to most offshore structures, except lighthouses and coastal engineering
installations, wave breaking is much more relevant for offshore wind energy applications
located typically in shallow waters. Breaking takes place when the water particle velocity
in the crest exceeds the celerity, the propagation speed of the wave. According to their
appearance, breakers are classified as spilling, plunging or surging (Figure 4.13).
Spilling breakers occur at a wave height of approximately 0.78 times the water depth in
shallow water or for 0.14 times the wave length in deep water. Seabed slopes and
strong winds in the direction of the waves reduce the breaking wave height further.
According to these conditions wave breaking begins to be of importance for water
depths below 2.5 times the significant wave height. In the southern North Sea this is the
case for depths below approximately 22.5 m which are typical for near future offshore
wind farms. The kinematics of spilling waves are well modelled by non-linear regular
wave theories for finite wave heights. Especially Stream function wave theory according
to Dean [4.40] is suited.
A nearly horizontal water surface and a spout form are observed at plunging breakers
resulting from sudden breaking due to a seabed slope. The description of the kinematics
and the pressure field is difficult and numerical models based upon computational fluid
dynamics rather than regular wave theories are required [4.42, 4.43]. As an alternative
to a modelling of the velocities in the crest other literature, e.g. DS 449, proposes a
distribution function for the impulsive wave forces in conjunction with a wave slamming
formula [4.44, 4.45]. A general overview is provided by [4.38]. For application to offshore
wind energy converters the reader is referred to [4.46].
Surging breakers are found when long waves of moderate height approach the shore
in very shallow water and if the sea bed slope is steep. They are not relevant for our
purposes.
67
Spectral density S(f)
Equidistant discreatisation
Aj
Ai
Ak
∆f
∆f = const. Ai ≠ Aj ≠ Ak
∆f
∆f
A
Frequency f
Equiamplitude discreatisation
Ai = A = const. ∆fi ≠ ∆fj ≠ ∆fk
A
A
∆fi
∆fj
Frequency f
∆fk
Enhanced discreatisation
Range 1:
Range 2:
Range 3:
equiamplitude discreatisation
as range 1, but refined spacing
around eigenfrequency fo
linearly increasing spacing
fo
Range 1
Range 2
Frequency f
Range 3
Figure 4.14: Spectral discretisation techniques for
simulation of random waves
Simulation of irregular waves
The wave energy spectra can be directly used for a linear spectral analysis of the sea
state loading as commonly done in the offshore technology. In many cases the nonlinear drag loading can be linearised and only if the system’s response is significantly
non-linear, a time domain approach is preferable.
Generation of time series with a random representations can be done by different
approaches. Generally the spectrum is discretised in equidistant frequency slices each
with the same width, ∆f, but with a different portion of the total energy and thus each
resulting in an elementary wave with different amplitude, Ai , (Figure 4.14 top). A number
of frequencies equal to a power of two is compatible with the Inverse Fast Fourier
Transform (IFFT). Usually sinusoidal elementary waves with random phase angles are
68
Chapter 4
superimposed. The summation of sine and cosine components with random amplitudes
is another technique, which is discussed with some other useful aspects in [4.33, 4.47,
4.48]. The time series from any of these techniques repeat with a period of 1/∆f. For
instance, 16384 = 214 frequencies are required to generate a three hours time series
without periodicity representing the spectrum up to a reasonable cut-off frequency of
0.628 Hz.7
Matthies et al. [4.46] used a spectral discretisation in slices of same energy but with nonequidistant frequency increments (Figure 4.14 middle) leading to an accurate
representation of the spectrum and some advantages for the simulation of extreme
waves. Furthermore, the time series is non-periodic, even with a relatively low number
of elementary waves of ,say, 100 to 200.
The resulting coarse frequency spacing in the high-frequent tail of the wave spectrum
is, however, problematical if the fundamental eigenfrequency is situated in this range,
which is typical for support structures of offshore wind energy converters. In the
DUWECS simulation code, extended during the present research, we overcame this
shortcoming by a specific discretisation of three spectral ranges (Figure 4.14 bottom).
First, in the vicinity of the eigenfrequency the spectrum is discretised by equi-energy
waves with a minimum frequency increment, ∆fmin = fo·ξo, related to both the
eigenfrequency, fo, and its damping ratio, ξo, [4.49]. This ensures a proper description
of the dynamic response behaviour. Next, the high-frequency tail of the spectrum is
divided up with a linearly increasing frequency spacing. Finally, the low-frequency part
of the wave spectrum, governing mainly the quasi-static response, is modelled by equienergy waves as in the original approach. Further details of the approach are
documented in [4.50].
Currents
Sea currents are characterised as near-surface currents generated by wind and waves,
sub-surface currents due to tides or thermosaline currents and near shore wave induced
surf currents. The UK Department of Energy and Germanischer Lloyd provide further
guidance [4.1, 4.13].
Due to the relatively shallow water depth, typical for a bottom-mounted OWEC, the
current loading is of lower significance but should be considered in the extreme load
calculations. Indirect effects such as erosion of the seabed in the vicinity of the structure
or wave-breaking due to an opposing current can also be relevant.
4.1.3
Other parameters
Variation of the sea level
The overall water depth at a location consists of a mean depth and a fluctuating
component which varies with the still water level (SWL) due to tides and storm surges
(Figure 4.15). The highest astronomical tide (HAT) and lowest astronomical tide (LAT)
are regular and predictable. Meteorologically generated storm surges are irregular and
are superimposed on the tidal variations. Total still water levels above HAT and below
LAT may occur. In addition, long-term changes of the sea level due to natural or manmade influences and motion of the sea bottom might occur. Again the UK Department
of Energy provides comprehensive guidance on the establishment of design values
7
This technique corresponds to the generation of the longitudinal wind turbulence in a single point of the
rotor plane (Section 4.1.1).
69
[4.1]. Water depth shown in
Admiralty charts are commonly
related to LAT or a Chart Datum
(CD), a reference level close to
LAT.
max SWL
positive surge
HAT
astronomical
tidal range
SWL
MSL
In shallow waters it is unclear
LAT
beforehand whether maximum or
CD
minimum water depth generates the negative surge
min SWL
most onerous loading. A lower
depth reduces the lever arm of the
Figure 4.15: Definition of water levels
hydrodynamic forces but increases
(The actual still water level (SWL) varies continuously
the wave steepness leading to more
depending on the actual tide and surge.)
non-linear behaviour and possibly
wave breaking. The more shallow the water the higher heave forces are acting on
support structures of the caisson type due to the pressure head of passing waves.
In a fatigue analysis the mean sea level (MSL) is often assumed as constant still water
level over the entire lifetime.
Icing and sea ice
Icing of the wind turbine and the superstructure from freezing snow, rain or fog can
occur but has lower importance than on land. More relevant seems the possible
accumulation of freezing sea spray, which was observed on platforms in the Dutch North
Sea and the Baltic. Norwegian standards state up to 80 mm ice thickness from wave
spray and a maximum of 10 mm from rain and snow for locations up to 560 N latitude.
Sea ice is a very significant extreme loading for Baltic sites due to a low salinity and cold
climate, whilst it is negligible for the North Sea or the Atlantic. Proper design
requires accurate investigations on the local ice properties, especially thickness and
Table 4.3: Indicative design properties of sea ice
Site
Thickness
tice [m]
Crushing
strength
σip [MPa]
Bending
strength
[MPa]
Northern Golf of Bothnia [4.51]
. 0.4 - 0.8
(max. 1.3)
1.0 - 3.0 †
not available
Rødsand, Denmark ‡ [4.52]
0.55
1.65
0.36
Øresund, Denmark ‡ [4.53]
0.57
1.9
0.5
design ice density 900 kg/m 3 for all sites
†
3.0 MPa intact moving ice at the coldest time of the winter
2.5 MPa intact slowly moving ice (e.g. thermal expansion), coldest time of the winter
1.5 MPa intact moving ice in the spring, temperature close to melting point
1.0 MPa partially weakened melting ice moving to melting temperature
‡
design value for return period 50 years
70
Chapter 4
strength, and the ice motion. Predictions are complicated due to variation of salinity, ice
temperature, ice history, size and velocity of ice floes and possible ice ride-up. Indicative
design ice properties are provided by Table 4.3 while ice force formulas are discussed
in Section 5.2.6.
Temperature
The large thermal capacity of the sea stabilizes the temperature variation known from
land. Differences between water and air temperature and temperature gradients are
important for determining atmospheric stability conditions, wind shear and turbulence
but are not directly relevant for structural design. Extreme minimum temperature values
for the selection of the steel and concrete grade range between -20o C in the southern
Baltic, -12o C in the Dutch North Sea to -4o C around the western British islands.
Maximum values relevant for the dimensioning of the wind turbine cooling system do not
exceed 25o C in the mentioned regions [4.1, 4.54, 4.55].
Marine growth
Fouling generates extra structural mass and additional hydrodynamic mass. The
thickness can grow up to 100 to 300 mm near the water surface. The modified surface
roughness increases the hydrodynamic drag coefficient (Table 5.4). For our type of
structures, mainly loaded by the hydrodynamic inertia force, the enlarged structural
diameter is of importance. Typical design properties are given by [4.1, 4.56].
Structures of the lattice or tripod type with slender members and greater surface are
more sensitive to marine growth than compact monopile or caisson designs.
4.2 Correlation of wind and wave parameters
So far we have addressed the met-ocean parameters separately, now we will draw our
attention to the correlation of wind and wave conditions, in particular.
Met-ocean databases and long-term statistics
The local site conditions such as fetch, water depth, sea bottom topography, etc. affect
also the correlation of the met-ocean parameters. Simple empirical closed-form
expressions are useful only for the early design stages (Section 10.1). Figure 4.16
compares the Neumann-Pierson relation [4.57] (10.1) valid for fully developed sea states
with the actual relation between the mean significant wave height and the mean wind
speed for different directions in the Dutch North Sea close to the site NL-1 (Figure 9.1).
The only alternative to site measurements is the application of numerical hindcasting
techniques that derive magnitudes and directions of simultaneous wind, wave, current,
tidal and surge parameters from meteorological data as atmospheric pressure
distributions and related wind fields. Only research institutes and meteorological offices
are capable of such complex models that process the data simultaneously at many grid
points of a large geographical region and for a long period covering the last 30 - 35
years. The results are validated against field measurements and are stored in
71
8
sig. wave height Hs [m]
extr. sig. wave height
H s50 = 6.9 m
NESS gridpoint 83,49
7
52^39'36" N Lat,
4^08'10" E Log
water depth 23 m (LAT)
distance from shore 31 km
6
5
4
wave direction NNW
deep water relation
(Neumann-Pierson)
3
omnidirectional
(averaged over
all directions)
2
wave direction SEE
1
0
0
5
10
15
20
25
30
hourly mean wind speed at hub height ( 58.5m ) [m/s]
Figure 4.16: Relation between mean significant wave height and
hourly mean wind at a Dutch North Sea site [4.59]
0
330
315
345 0.12
15
0.10
NESS gridpont 83,49
30
45
0.08
300
52^39'36" N Lat, 4^08'10" E Log
0.04
285
3 h mean wave direction
75
0.02
270
3 h mean wind direction
60
0.06
N
90
0.00
255
105
240
120
225
210
195
165
E
W
135
150
S
180
Figure 4.17: Distribution of mean wind and wave direction
at a Dutch North Sea site [4.59]
30%
probability [-]
25%
NESS gridpoint 83,49
52^39'36" N Lat, 4^08'10" E Log
20%
15%
10%
5%
180
150
120
90
60
30
0
-30
-60
-90
-120
-150
-180
0%
misalignment of wave direction w.r.t. wind direction [deg]
(15 degree sectors)
Figure 4.18: Misalignment of mean wave direction with respect to
mean wind direction at a Dutch North Sea site [4.59]
72
Chapter 4
high-quality met-ocean databases [4.58]. Beyond a water depth of about 20 m such
environmental data can be derived without too much disturbance by the local bottom
topography. Application to sites with lower depth and considerable changes of the depth
and bottom topography relative to the nearest grid point, should be carried out with care.
For Northwestern Europe the NESS database (North European Storm Study) is of
particular importance [4.59, 4.60]. Within the Opti-OWECS study, data for two grid points
in the Dutch North Sea were made available though Shell International Exploitation and
Production B.V. A continuous record of nine years enabled reliable fatigue calculations
and extreme parameters could be extrapolated from all storms during a period of 30
years. In Chapters 9 and 10 these grid points are matched with the sites NL-1 and NL-5
(Figure 9.1).
As an example, Figure 4.17 compares the distribution of the mean wind and mean wave
direction showing prevailing southwesterly winds. Two main wave directions are
observed which correspond to the main wind direction and the northern direction from
the Central North Sea with associated long fetch. In Section 9.6.1 the wave
misalignment is further investigated and collinearity of wind and waves is found a
reasonable assumption for onmidirectional waves as 65% of the waves are within a ±
22.5% sector around the wind direction (Figure 4.18). The three-dimensional scatter of
wind, wave height and wave period for the same site is shown in Figure 6.7 and Table
8.1.
Short-term statistics
It is common practice to assume the met-ocean conditions as statistically constant over
an averaging period of three hours. In wind engineering, often a period of 10 minutes or
one hour is used instead. Both periods are situated in the spectral gap of winds and
waves (Figures 4.2 and 4.8). In the following the entire set of environmental conditions,
rather than only the wave parameters, is denoted as a sea state.
Fatigue analysis is an in principle straight-forward but cumbersome task if a suitable
model of the entire offshore wind energy converter (Chapters 5 and 8) and a scatter
diagram with all sea state parameters is available. Both time and frequency domain
approaches rest upon the independence of wind and wave loading on short time scales.
Establishment of the simultaneous extreme environmental parameters for a certain sea
state is not easy for several reasons. Wind and wave loads can have a similar order of
magnitude. In contrast, the design of most bottom-mounted offshore platforms is
governed by extreme wave loads, whilst the wind contribution accounts for typically
around 10% of the total loading. Secondly, dynamic response is more pronounced than
for most other structures of the petroleum industry. Finally, the operational conditions
of the OWEC require attention. The extreme response can occur during the power
production mode associated with sea states far below the extreme 50 years conditions.
Joint probabilities for extreme gust and extreme design waves can be derived assuming
Gaussian wind turbulence and Rayleigh distributed waves. On this rationale
Germanischer Lloyd recommends extreme load cases associated with either the most
probable extreme wave or wind gust for a given sea state (Table 4.4). Such treatment
is suited to engineering purposes and is followed within this thesis but it has
shortcomings. Two thirds of the maxima are higher than the most probable extreme
value and only a probabilistic description is accurate. The most onerous environmental
73
Table 4.4: Wind and wave correlation for extreme load cases according to
Germanischer Lloyd [4.13]
OWEC
condition
production
Identifier
E1.1
E1.5 ‡
E2.1
parked
E2.2
E2.4
†
‡
Hydrodynamic conditions
Wind conditions †
mean wind within
production range:
- extreme operating gust
13 m/s
- reduced 1 min gust
sea state corresponding to mean
wind:
- reduced wave Hred = 1.32 Hs
50-year mean wind:
- extreme 3 s gust
(1.2 times mean wind)
- reduced 1 min gust
(1.09 times mean wind)
50-year extreme sea state:
- reduced wave Hred = 1.32 Hs50
50-year mean wind
50-year sea ice
- extreme wave Hmax = 1.86 Hs
- extreme wave Hmax = 1.86 Hs50
collinear mean wind and wave direction, but prescribed change in instantaneous wind direction,
generally the E1.5 load case is not governing
parameters should be derived from the parent distribution of the extreme response
rather than the joint-probabilities of the environmental conditions. Recently the
application of structural reliability methods, developed in advanced offshore engineering
to offshore wind energy converters proposes solutions for this problem [4.61].
Safety factors
Germanischer Lloyd proposed a lower partial safety factor on extreme environmental
loads of γF = 1.35 for offshore wind energy application instead of the value of γF = 1.5 for
land-based installations (Table 4.2). This is due to the loose correlation between shortterm extreme wind and waves loads and the reduced risk of human life due to a major
failure of an OWEC compared to a wind turbine on land or an offshore structure of the
petroleum industry.
4.3 Soil conditions and auxiliary external effects
Soil conditions
The demands of OWEC foundations differ from those of more common offshore
structures.
Governing design criteria for the foundation of an oil platform are often the ultimate
bearing capacity and sliding resistance, as well as overall stability against overturning.
Therefore the minimum strength under static (or dynamic) loading is one of the most
important soil parameters. Mainly the lower bound of the foundation stiffness is crucial
since it corresponds with the minimum fundamental eigenfrequency and maximum
dynamic amplification of the wave loading. Any variation of the soil properties leading
to a stiffer behaviour of the structure and hence less onerous condition is welcome.
74
Chapter 4
In contrast, both minimum and maximum soil stiffness are relevant for OWEC since the
structural eigenfrequencies have to be tailored with respect to the three main excitation
ranges, the dynamic wave loading and both rotor and blade excitations.
Chapter 10 proposes a rational treatment of the uncertainties in the foundation
behaviour during the course of the design process.
The values of the soil parameters for strength and stiffness vary significantly depending
on soil type, loading history and location. Tables 4.5 and 4.6 give indicative soil
properties that correspond to the models described in Sections 5.2.2 and 5.2.3.
Table 4.5: Indicative soil properties for gravity foundations
Submerged unit
weight γsub
[kN/m3] [4.62]
Dynamic shear
modulus Gdyn [MPa]
[4.63] †
sand (loose)
sand (medium dense)
gravel (sandy, dense)
10
11
10
50 - 120
70 -170
100 - 300
mud, clay
clay (soft to firm)
clay (stiff to very stiff)
9
10
11
3 - 10
20 - 50
80 - 300
Soil
rock (layered, cracked)
rock (solid)
Poisson
ratio ν
[-] [4.63] ‡
,
* 0.25 - 0.35
-
0.35 - 0.45 ††
0.45 - 0.49 ‡‡
1,000 - 5,000
4,000 - 20,000
†
[4.63] recommends calculation with ± 50% variations to account for uncertainties
[4.38] recommends the undrained Poisson ratio ν = 0.5 for cyclic loading
††
‡‡
silt, depending on sand and clay content,
clay, depending on water content
‡
Table 4.6: Indicative soil properties for piled foundations [4.38, 4.64]
Sand / gravel
silty, clayed
loose
medium dense
dense
dense, very dense
Clay
clay (soft to firm)
clay (stiff to very stiff)
Submerged unit
weight γsub
[kN/m3]
Friction
angle N
[deg]
Initial subgrade
reaction K
[MPa/m]
7 - 13
7-9
8 - 10
10 - 11
11- 13
25 - 35
28 - 30
30 - 36
36 - 40
40 - 45
2-7
7 - 28
28 - 41
41 - 59
Submerged unit
weight γsub
[kN/m3]
Undrained
shear strength
Cu [kPa]
Strain at 50% of
peak stress ε50 [%o]
4-7
9 - 10
10 - 50
50 - 200
5 - 20
5 - 20
75
Scour
Loose sand and soft clay are sensitive to seabed currents around the base of the
structure. Strong tidal currents and wave induced currents in the breaking wave zone
may lead to sediment scour, which will affect the generic foundation types differently.
Piled foundations will suffer from a localised reduction in overburden pressure and
a loss of lateral resistance. The longer lever arm of the environmental forces directly
increases the bending loads in the pile and a softening of the dynamic characteristics
results in more pronounced dynamic response on wave excitation and possible
resonance with the rotor excitation for soft-stiff designs. The magnitude of scour for large
monopiles in shallow water is difficult to predict but values of three to six metres seem
realistic for silicious soils. In the design of offshore jacket structures commonly 1.5 times
the leg diameter of local scour and one metre of global scour is allowed for [4.38].
Germanischer Lloyd recommends a value of 2.5 times the pile diameter if no other data
are available [4.13]. This figure seems, however, too conservative for large monopiles.
Gravity structures may undergo erosion of soil from underneath the base of the structure
if skirts do not penetrate until sufficient depth. In shallow waters scour is likely to exceed
the maximum value of 2 m observed at structures in the central and northern North Sea
[4.38].
There are two main design options to deal with sea bed erosion; allow for scour in the
design or to monitor the actual scour and replace the material if necessary. The latter
is also needed if scour protection was already provided during installation.
Auxiliary external effects
Corrosion is an important consideration in the design of both steel and concrete marine
structures. Sites in the North Sea and Atlantic Ocean are more sensitive than in the
Baltic Sea owing to the harsh environment and the 4 to 6 times higher content of
aggressive ions [4.65].
Little evidence exists about the frequency of lightning strikes at offshore wind energy
converters compared to their land-based counterparts. Generally a higher degree of
lightning protection is recommended due to greater exposure of the structure, the longer
downtime after a severe incident and the potentially higher repair cost.
Depending on the site and design conditions, other loadings such as ship impact,
earthquake, mobile sand features, etc. might require particular consideration. Initial
inclination or settlement during exploitation can induce significant gravity loads in the
slender and tall support structure and might affect proper function of the yaw system of
the wind turbine.
Influences arising from the energy consumers need attention during design of an
offshore wind farm, as well. Grid failure and/or loss of load are to be considered as
normal external conditions while major frequency, voltage and load fluctuations,
interfering voltages and line short-circuit are examples of extreme influences which can
affect not only the electrical system and the drive train but also the rotor and support
structure.
76
Chapter 4
Modelling of Offshore Wind Energy Converters
77
CHAPTER 5
MODELLING OF OFFSHORE WIND ENERGY CONVERTERS
Design calculations for offshore wind energy converters (OWEC) require a mathematical
model which is based upon the state of the art in both wind energy and offshore
technology. In order to limit the complexity to a level appropriate to engineering
application a particular approach was developed emphasising aspects most relevant to
bottom-mounted offshore wind energy converters.
In this chapter the modelling of the two main sub-systems, the wind turbine and the
support structure is elaborated in Sections 5.1 and 5.2. There are however several subsystem aspects that affect the entire system. So, Section 5.3 discusses the time
invariant properties of the wind turbine rotor, the aerodynamic damping and the variation
of the system dynamics from an integral viewpoint.
Obviously the choice of a suitable analysis approach for the entire system must be
compatible with the properties and modelling of all sub-systems. Therefore this chapter
describes only the treatment of the sub-systems, while Chapter 6 introduces the socalled integrated, modular, non-linear time domain simulation approach for the entire
offshore wind energy converter.
Aiming at a compact description, most attention is given to those aspects of offshore
wind energy converters which differ from their land based counterparts.
5.1 Wind turbine modelling
General remarks on the applied model
For this thesis the onshore wind turbine design tool DUWECS (Delft University Wind
Energy Converter Simulation) [5.1, 5.2] was extended by many features required to
analyse offshore wind energy converters. Emphasis was put on an integrated model
including wind turbine and support structure. So most enhancements were required with
respect to the support structure and its loading (Section 5.2) and the computational
efficiency of simulations of the simultaneous aerodynamic and hydrodynamic loading
(Sections 6.3 and 6.4). The wind field representation and the structural dynamics of the
wind turbine model [5.3] was extended but the implementation of the rotor aerodynamics
remained nearly unchanged [5.4].
As a matter of fact the applied DUWECS model includes (Figure 5.1) all major features
required for wind turbine design calculations but suffers some limitations with respect
to the recent developments of onshore design calculations. Shortcomings include the
78
Chapter 5
flapwise deflection
V
flap & lead-lag hinge,
(pitch rotation)
ux ,(uy )
Ω
hinge
eccentricity
lead-lag
deflection
azimuth angle
Ω•t
Figure 5.1: Schematic representation of the applied wind turbine model [5.3]
(Options for third blade and lateral motion not shown)
three-dimensional turbulence, state of the art rotor aerodynamics and multiple blade
bending modes. Table 5.1 compares the used program to a popular, commercial
simulation code [5.5]. Due to the consistency with the work within the Opti-OWECS
study most simulations in this thesis are computed with the ROTOR2 module of
DUWECS, a two-bladed rotor model with fore-aft translation degree of freedom. For
some particular analyses the more recent module ROTOR3 with two or three rotor
blades, both fore-aft and lateral translation degree of freedom and some other
improvements are used. The overall suitability of the chosen modelling was confirmed
with a benchmark experiment carried out by comparison to another commercial design
tool [5.6, 5.7].
The rest of this sections presents the essentials of wind turbine modelling for design
calculations of OWEC more from a general point of view. Comments on the specific
implementation in this thesis are given where appropriate. More comprehensive
information on the varieties of wind turbine modelling is available from overview papers
[5.8, 5.9] and textbooks [5.10 - 5.14].
5.1.1
Non-homogeneities of the wind field
The inflow comprises two contributions; a stochastic part from the ambient and wind
farm induced turbulence and a deterministic portion owing to wind shear, yawed wind
direction and tower influence. The former aspect was discussed in detail in Section 4.1.1
where the importance of a sophisticated representation of spatial and temporal structure
of the wind was highlighted.
Now we will have a closer look at the stationary composition of the wind field. Modelling
of the stationary flow field is straightforward. A geometric transformation of the flow
accounts for yawed flow due to a misalignment or delay of the yaw tracking system and
inclined stream owing to the tilt angle of the rotor shaft, unstable atmospheric conditions
or complex terrain conditions. For design calculations a power law for the wind shear
rather than a logarithmic wind profile is used.
Similar to the rotational sampling of turbulence, also the deterministic disturbances of
the spatial flow field are observed as temporal variations of the local wind speed and
79
15 deg yawed flow
5 deg tilt angle
0.11 wind shear exponent
50
40
60
50
nacelle tilt moment [kNm]
blade flap moment [kNm]
60
40
30
30
20
20
2 blades
10
2 blades
10
0
0
-10
-10
3 blades
3 blades
-20
-20
-30
-30
1
2
3
4
5
6
7 8
order
1
9
2
3
4
5
6
7
8
9
Figure 5.2: Effect of yawed flow, wind shear and rotor tilt on Fourier Series
decomposition of the blade bending and nacelle tilt moment
(Rigid model of a two- or three-bladed, stall controlled 500 kW
turbine with constant solidity and tip speed)
25
25
tower shadow,
5 deg tilt angle
20
15
10
2 blades
5
0
-5
nacelle tilt moment [kNm]
blade flap moment [kNm]
20
15
10
2 blades
5
0
-5
3 blades
-10
3 blades
-10
1
2
3
4
5
6
7 8 9
order
1
2
3
4
5
6
7
8
Figure 5.3 Effect of tower shadow and rotor tilt on Fourier Series
decomposition of the blade bending and nacelle tilt moment
(Rigid model of a two- or three-bladed, stall controlled 500 kW
turbine with constant solidity and tip speed)
9
• tower & foundation
Structural dynamics
• blade
• drive train
• nacelle
• tower top
• unsteady profile
aerodynamics
Chapter 5
wind shear (power law)
tower shadow (dipole)
yawed and inclined flow
axial turbulence (SWING 3 code, rotating
observer) with deterministic variation of lateral
component by sine law
• fore-aft (ROTOR2),
fore-aft and lateral (ROTOR3)
• Finite Element model (n modes)
• flap and lead-lag hinge model
• rotation, torsion
uniform axial induction
cL and cd from tables
Johnson correction
lift loss factor
depending on airfoil tables, e.g. Viterna &
Corrigan, Snel, etc.
• 1st order differential equation on induction
factor
• ONERA (not used)
•
•
•
•
•
•
•
•
•
DUWECS
axial and tangential induction per annuli
cL, cd and cm interpolated from tables
Glauert correction
Prandtl tip loss factor
depending on airfoil tables
wind shear (power law)
tower shadow (dipole)
yawed and inclined flow
3D turbulence (Veers model)
2 flapwise, 2 edgewise modes
rotation, x- & y-bending, torsion
tilt
fore-aft, lateral & torsion
• 1st fore-aft, lateral & torsion mode
•
•
•
•
• dynamic wake model for induced velocity per annuli
• Stig Øye dynamic stall model
•
•
•
•
•
•
•
•
•
FLEX4
Table 5.1: Comparison of applied wind turbine modelling to a state-of-the-art simulation code
Rotor aerodynamics
• blade element momentum
theory
• turbulent wake state
• blade root and tip effects
• 3D correction of profile
aerodynamics
• dynamic inflow
• stochastic
Wind description
• deterministic
Feature
80
81
•
•
•
•
• state space analog controller
• digital controller and filters
• link of user built control routines incl. transducer
and actuator dynamics
• linearisation option
Transmission & Generator
Control
gear box flexibility (optional)
induction or synchronous generator
fixed or variable speed
link of user built transmission and generator
routines
DUWECS
Feature
• link of user built routine for transducer and
actuator dynamics
• link of user built control routines
• induction or synchronous generator
• fixed or variable speed
• link of user built generator routines
FLEX4
Table 5.1 (cont.): Comparison of applied wind turbine modelling to a state-of-the-art simulation code
82
Chapter 5
direction in the rotating reference frame. Yaw and wind shear are experienced by the
rotor blades as nearly perfect sinusoidal excitation with the rotor frequency (1P).
Figure 5.2 illustrates a Fourier decomposition of the quasi-static blade root and nacelle
tilt moments of two fictitious turbines with same solidity. The blade root loading of a
blade of a three-bladed rotor is about 67% of a two-bladed design owing to the reduced
blade area. In the fixed reference system of the nacelle the excitations are seen mainly
as multiples of the blade passing frequency. While a significant averaging between the
three blades is observed, the opposite holds for the two-bladed case where the tilt
contributions from both blades are acting in phase.
For upwind orientated turbines the blocking effect of the tower and nacelle is commonly
modelled by a solution of the potential flow problem. The blade suffers the sharp velocity
deficit and deformed streamlines in front of the tower mainly as a 1P excitation but a
large number of higher harmonics contribute to a slowly decreasing but low level (Figure
5.3). In the fixed reference frame of the nacelle the forcing by the blade passing
frequency is rather high in the two blades case. This is the reason why sometimes twobladed rotors are combined with a teeter hub which is compensating the 2P loading by
a counteracting out-of-plane motion of the blades. The three-bladed rotor is behaving
again similar to a disk which smoothens the inflow. It is amazing to see that the three
blades are sampling also a low 2P excitation.
Two-bladed concepts are appealing for offshore siting due to various aspects. On sea
they can operate with an optimal and high tip speed ratio. Installation is simplified and
aesthetics are of lower concern. Nonetheless, as seen at the above mentioned example,
the dynamic loads of two-bladed turbines are considerably higher (at least for constant
tip speed). Moreover the periodic variation of the system properties results in a more
complex dynamic behaviour.
5.1.2
Rotor aerodynamics
Blade element momentum theory
All wind turbine design tools calculate the aerodynamic loads with a combination of
blade element and momentum theory. In order to overcome the shortcomings of this
rather crude method several engineering rules are applied as corrections. Despite the
rapidly increasing computer power, more advanced methods applying Computational
Fluid Dynamics [5.9, 5.15] are still under development and will most likely only be able
to support and validate much simpler design tools for years.
The flow effects and associated models in rotor aerodynamics are often distinguished
by the different spatial and temporal scales involved (Figure 5.4). The global flow field
has dimensions in the order of the rotor diameter and a time scale of several seconds
while the blade flow field is related to the chord length and periods of 10 to 50
milliseconds which a particle needs to cross the blade.
The global flow field is described by Froude’s actuator disk theory originating from 1889
[5.16] and was adopted to wind turbine applications by Lancaster [5.17] and Betz [5.18].
Commonly it is corrected for conditions with high axial loading (turbulent wake state),
for instance with empirical relations according to Glauert [5.19]. Blade element or wind
theory with empirically determined airfoil characteristics expresses the local lift and drag
forces [5.20].
Both flow fields are linked by the axial (out-of-plane) and tangential (in-plane) ‘braking
action’ at the rotor disk, denoted as induction. On the one hand the induction determines
83
dFx
V(1-a)
dLift
α
V
dFx
dDrag
W
Figure 5.4: Blade Element Momentum theory linking of the global rotor flow field
(left) to the local blade field (right) by the induction force dFx of an annulus
the velocities at the rotor disk by the global momentum balance. On the other hand the
local aerodynamic forces are depending on the magnitude and angle of incidence, α,
of the apparent wind velocity composed from four contributions: undisturbed wind
speed, V, including turbulent and deterministic variations; blade azimuthal velocity, Ω·r;
blade motion due to deformation of the blade, nacelle and support structure and the
axial and tangential induced velocities1 (Figure 5.4).
The swept area is divided into 10 to 20 annuli under the rough assumption that no
spanwise interference occurs and the induction is distributed uniformly over all annuli.
The axial and tangential inductions are determined from the static or dynamic
equilibrium depending on the used wake model. Obviously the induction is nonhomogeneous if only two or three blades operate. So a correction according to Prandtl
[5.21] is applied, denoted with the misleading name ‘tip correction’. The Blade Element
Momentum (BEM) method was applied for wind turbine rotor design by Schmitz [5.22]
and has been synthesised for modern computational use by Wilson and Lissaman et
al. [5.23] in the 1970's. Since then it is a daily tool for engineers using it successfully but
also a topic for researchers discussing it frequently.
Semi-empirical corrections
The Blade Element Momentum approach has been continuously improved by various
engineering corrections. Reference [5.24] provides a comprehensive overview.
The aerodynamic coefficients of the thick airfoils at the blade root sections determined
from two dimensional wind tunnel experiments are now corrected for three dimensional
and rotational effects of the separated flow where stall is delayed and up to 50% higher
lift is observed than predicted by the elementary theory (Figure 5.5 left) [5.25].
In the previous section we got an impression of the high load fluctuations that are
1
The axial and tangential induced velocities, a·V and a’·V, are related to the undisturbed (axial) velocity,
V, and circumferential velocity, Ω·r, by the non-dimensional induction factors a and a’. The tangential
induction of a modern wind turbine with high tip speed ratio is very small.
84
Chapter 5
2
cL
cL
3-dimensional
& rotating
1.2
C
1.6
unsteady
B
1.2
A
D
2-dimensional
quasi-steady
10o
attached flow
Left:
Right:
25o
α
α
10o
20o
separated (stalled) flow
Figure 5.5: Extensions of elementary rotor aerodynamics:
Three-dimensional and rotating effects delay the stall.
An oscillations of the angle of attack, α, result in a hysteresis of the lift
coefficient, cL. (A: Flow reversals within boundary layer, vortex formation.
B: Vortex detaches and moves over surface.
C: Vortex passes trailing edge, full stall. D: Flow re-attachment.)
caused by yawed flow. The stream tube concept of the BEM approach cannot
sufficiently describe larger misalignment of rotor axis and flow pattern since undisturbed
wind and the wake are no longer in one line. One reasonably successful modification
corrects the ‘mean’ induction calculated by standard BEM by a rotor azimuth dependent
induction [5.26].
The flow around wind turbines operating in the atmospheric boundary layer is inherently
unsteady. Dynamic inflow denotes all unsteady aerodynamics associated with the rotor
flow field. A common engineering approach changes the static equilibrium of the
induction into a radially dependent, dynamic relation with first order delay characteristics
[5.27].
On the blade scale the fluctuating external conditions result in hysteresis effects of the
aerodynamic coefficients in the stall region (dynamic stall), depending on the blade
geometry, the frequency and the amplitude of the flow variation. Figure 5.5 (right)
presents a typical behaviour. Common models from helicopter aerodynamics, as for
instance reviewed and extended by Björck et al. [5.28], express the time derivative of
the lift coefficient as a function of the instantaneous value, the velocity and the
acceleration of the angle of incidence .
Analyses of the aero-elastic stability, especially important for large, flexible rotors
operating in stall, require both a sophisticated model of the structural dynamics and the
unsteady profile aerodynamics [5.29].
85
5.1.3
Structural dynamics
It is now state of the art to consider dynamic loads in the design process of all load
carrying major components of a wind turbine. Therefore design tools comprise models
of the structural dynamics of rotor blades, hub, drive train including support and brakes,
generator, yaw system, tower, foundation and controller. Due to the specific combination
of a rotating structure undergoing gross overall motion with a fixed support structure,
particular mathematical models are required which were until quite recently not available
in general purpose codes. Therefore most tools were developed over a long period by
research institutes particularly with respect to the specific needs of the wind turbine
community.
Such models are using either the Finite Element representation or the modal analysis.
While the former requires extensive computer power during the actual time domain
simulation process the latter approach involves the cumbersome derivation of the
equations of motion during the establishment of the model. The majority of tools follow
the modal analysis with eigenmodes and frequencies extracted from a multi-body model
(Figure 5.1) or a Finite Element model. The typical number of degrees of freedom of the
dynamic models varies between 20 and 40 describing the relevant dynamics up to 5 or
10 Hz. Due to the non-linear kinematics and the non-linear rotor aerodynamics all
commercial tools rest on time domain Monte-Carlo simulations.
In Sections 6.1.2 and 6.2 we introduce a suitable analysis approach in more detail since
the support structure of an offshore wind energy converter features distinctly different
modelling requirements than the wind turbine part (Table 6.1).
5.1.4
Mechanical-electrical conversion system
The importance of a transmission and generator system for the system dynamics is
explained by two examples of modern turbine designs.
The massive primary shaft of a common 1.5 MW wind turbine has a diameter of
approximately 600 mm and a length of some 2.5 to 3.5 metres. Nonetheless it is beside
the slender blades and tower one of the most flexible and oscillatory components owing
to the enormous torque and the inertia of the rotor and generator.
Variable speed designs require a frequency converter with power electronics for at least
a portion of the generator power. Such generator - converter systems are highly active
and can be used for tailoring energy yield, fatigue loads, braking and damping the
turbine. They have less in common with the traditional, robust induction generators
modelled simply with a linear torque - rotor speed characteristic.
Only the first one or two torsion modes of the drive train are of importance. Therefore
relatively simple models are surviving. Drive train models with one or two degrees of
freedom, for the low speed or both low speed and high speed shaft flexibility, and user
defined generator routines are implemented in design tools. The three or four stage gear
box (if any) is commonly represented solely by a rigid, massless transmission, the
dynamic properties of which are often unknown to both the turbine and gear designer.
In contrast to the generator manufacturer, only the low-frequency dynamics of the
generator or generator - converter system are relevant for the turbine designer. Models
for the various used types, e.g. induction generator, double-fed asynchronous
generator, synchronous generator with DC-link, are described in references [5.1, 5.30 5.32].
86
Chapter 5
5.1.5
Supervisory and power control
Control principles applied at wind turbines
The technical and economic performance depends on proper understanding of the
control system and its interaction with the rest of the turbine.
Actions of the supervisory control as start, stop, brake, yaw, synchronisation manoevers
have to be considered in design calculations because they contribute to the
accumulated fatigue and have an important impact on extreme loads related to design
gusts or emergency actions. Commonly the reaction of the simulation model is observed
with respect to synthesized signals, of for instance pitch variation, rotor speed ramp or
brake torque, or triggered occurrences as emergency shut-down and power loss.
Dynamic power control has to meet three basic requirements: safe limitation of the
elementary machine loads of power, rotor speed and thrust, high energy yield at partial
load conditions and moderate fatigue loads by limitation of load variations and control
actions. Stability, good disturbance rejection and robustness must be met under the
variable operational conditions and with autonomous operation. Different answers are
provided by the three currently popular wind turbine concepts of either traditional stall fixed speed, active stall - fixed speed or variable pitch - variable speed.
At traditional stall controlled turbines these tasks rest entirely on the aerodynamic
behaviour of which we still have an incomplete understanding. An induction generator
with a high turn-over torque ensures fixed rotor speed which results in large angle of
incidence and stall when the wind speed increases. The reduction of lift and increase
of the drag are limiting the rotor torque and power. Apart from the rotor aerodynamics
and possible aero-elastic effects modelling is straightforward.
Active stall designs follow the same principle because the stall effect itself is faster and
theoretically simpler than any mechanical control. Relatively slow variations of the blade
pitch angle in the rotor plane (to larger angles of attack) assist the stall effect and adjust
it to the mean wind speed and local site conditions. Furthermore, start and stop action
and parking is simplified and extreme loads are reduced. Simple single-input - singleoutput (SISO) controllers of PI or PID type with moderate time constants are sufficient.
An opposite pitch variation out of the rotor plane is used at the variable pitch (or pitch
controlled) concepts to reduce the angle of attack and accordingly both lift and drag
forces. During the early times of the industry, when power electronics were not present
or too costly, the principle was combined with almost constant rotor speed. Later it was
realised that pure pitch control with pitch rates up to 10 o/s and higher was not fast
enough to counter the turbulence induced fatigue and power variations. Therefore
turbines of intermediate size are using a successor of the classical pitch principle. A
limited variability of the rotor speed is facilitated by increased slip of the asynchronous
generator. Optimum performance by high technological effort is achieved by the
variable-speed, variable-pitch concept that gains increasingly more contenders within
the community in recent years.
Variable speed enables fast load control through the entire range between cut-in and
cut-out wind speed as the rotor is acting as flywheel flattening the fluctuating power in
the wind. In the partial load regime noise and energy yield are optimised as well. Pitch
control with moderate pitch rates of about 2 to 5 o/s is commonly activated only above
the rated wind speed to limit the power. High pitch rates are required only at emergency
stops. Obviously, control of variable pitch - variable speed machines is complex and
87
global inputs
(controlled)
Controller
global
outputs
1
Wind
global inputs
(uncontrolled)
V
S G
MR
Rotor
S
Transmission
MG
G
S G
Generator
IG
G
UG
G
Grid
Fx , etc.
dux/dt
interaction
variables
Tower
Wind Energy Converter
Figure 5.6: Modular structure of the onshore version of the DUWECS code
(Rotor, tower, transmission and generator modules coupled by interaction variables. Global
inputs and outputs link the wind, grid and controller to the plant.)
interrelated with the dynamics of the entire system. At partial load a single-input - singleoutput controller of PID or higher order can be applied. Both speed and power serve
inputs at rated power while pitch and a converter - generator quantity are controlled
(multi-input - multi-output system, MIMO). Balancing the above mentioned control goals
requires a combination of different controllers which are switched close to the rated wind
speed [5.33] and the parameters of which are adjusted gradually to the operating
conditions [5.1, 5.34].
Implementation of control
The DUWECS code used within this thesis was designed especially for analysis of
controlled wind turbines. Analog or digital controller and filter modules can be easily
implemented in the modular structure of the program. Figure 5.6 shows a model with a
simple SISO controller of PI type as used later for the pitch controlled WTS 80 turbine.
A first order response behaviour of the actuator was included in the controller as well.
More advanced modelling of pitch systems is described in reference [5.35].
The simulation code DUWECS is able to derive a linear state space model of the entire
OWEC (5.1) with respect to deviations from a point of equilibrium. This linearisation
option facilitates application of classical linear control theory during the design of the
controller. In addition the system eigenfrequencies and damping
can be determined from the complex eigenvalues of the state matrix, A.
x0
y
'
'
A x
C x
% Bu u
% Du u
%
%
Bv v
Dv v
(5.1)
where:
x state vector
u controlled input, e.g. pitch angle
ν external input, e.g. wind speed
y output, e.g. rotor speed
A state matrix
C output matrix
Bu , Bν input matrices
Du , Dν feed-through matrices
(x, u, v and y are defined with respect to their equilibrium values. Du vanishes for most
real systems.)
88
Chapter 5
The state space model itself is derived by treating the wind turbine as a black box and
observing the variation of the outputs and the derivatives of the states on successive
perturbations of the state and the input variables.
Suppose the non-linear wind turbine model at an operating point x , u , v is given by
the vector functions x0 ' F x , u , v and y ' H x , u , v then an element
A(i,j) of the state matrix A is determined by the finite central difference.
Ai j x , u , v
F i x , u , v x ' x % ∆ x & Fi x , u , v x ' x & ∆ x
j
j
/ j j
/ j j
'
2 ∆ xj
(5.2)
If an analog controller of the wind turbine is defined in state space form (5.3) is applied,
an extended state matrix, Aclosed-loop, according to (5.4) can be established.
x0 c ' Ac xc % B c y % R c r
u ' Cc xc % D c y % S c r
(5.3)
where:
y controller input, i.e. turbine output
xc controller state vector
u controller output, i.e. turbine input
r reference signal
Ac , Bc , Cc, Dc, Rc, Sc state space matrices of the controller
Aclosed&&loop '
A % Bu D c C
Bu ( C c % Dc D u Cc )
Bc C
Ac
(5.4)
In Section 7.1.3, Eqn. (5.4) is used to calculate the aerodynamic damping of the support
structure including the important dynamic effects of the control system.
5.2
Support structure
Tower and foundation should be considered as one support structure sub-system. This
is meaningful with respect to the modelling but also offers a potentially cheaper design
solution. In most aspects the treatment of the support structure is comparable to
common offshore platforms from the oil and gas industry. The main differences lie in the
relatively small size, the lower water depth and the more dominant aerodynamic loading.
For further details not covered in this section one is referred to textbooks [5.36, 5.37].
5.2.1
General aspects of structural modelling
Choice of Finite Element representation
The diversity of generic types of support structures, e.g. monotower, braced or lattice
tower configurations either piled or gravity (Figure 3.16), requires equally diverse
modelling features that can be achieved only by Finite Element analysis. For our
purposes beam element models are sufficient.
Within this thesis and in contrast to both the previous work of the author [5.38] and other
recent OWEC design tools, the Finite Element representation is not directly
89
implemented in the design tool. Instead the general purpose code ANSYS™ 5.3 [5.39]
with a consistent formulation of mass and stiffness matrices is used as a post-processor
for the generation of the support structure model. This choice, further justified in Section
6.1.2, enables the treatment of support structures of ‘arbitrary’ complexity rather than
only monopile or monotower structures. In addition, specific aspects are available as
submerged and hydrodynamically loaded elements, stress stiffening, non-linear
structural characteristics, various dynamic analyses capabilities and techniques for the
reduction of degrees of freedom
Prior to the separate consideration of piled and gravity structures some general remarks
on mass and damping aspects relevant for both generic types are given.
Mass aspects common to all support structure types
A structure moving in a calm sea or in a sea state experiences a hydrodynamic inertia
force which is proportional to the structural acceleration. This force can be considered
as a part of the fluid loading but it is more conveniently computed by superimposing the
so-called ‘water added mass’ on the structural mass (Section 5.2.5) .
The interior of offshore structures below the sea surface is often filled up with water to
counter the buoyancy force and the external hydrostatic pressure. In a dynamic analysis
this mass has to be added to the structural mass analogous to the hydrodynamic added
mas.
The mass of the marine growth of space frame structures can be considerable and
should be taken into account.
Damping aspects common to all support structure types
Four sources of damping are important for an OWEC, which are given in the order of
decreasing relevance:
• aerodynamic damping
• structural damping
• soil damping
• hydrodynamic damping
The aerodynamic damping is a system rather than sub-system property. It is related to
the interaction between the support structure and the wind turbine and is discussed with
the modelling of other system properties in Section 5.3.
Internal material friction and bolted or welded joints generate structural damping which
is conveniently considered as modal damping. Typical values recommended for
dynamic response analysis of wind loaded structures are given in Table 5.2. In this
thesis generally 0.5% of material damping is considered for steel structures. Combined
structural and soil damping of wind turbine towers in the range of 0.2 to 0.5% has been
measured by Schaumann and Seidel [5.40].
The different types of soil damping are elaborated in conjunction with the gravity and
piled foundations.
Finally the structural motion in the water leads to hydrodynamic damping. Radiation of
vibration energy is very low, but damping due to the viscous drag force on flexible,
submerged members as cables and risers is well known in offshore engineering. For our
type of structures, which are relatively stiff below the water surface, the effect is,
however, negligible (Sections 5.2.5 and 9.4).
90
Chapter 5
Table 5.2: Recommended structural damping ratios [5.36]
Structure
Modal damping ratio
(1st bending mode, as percentage of
critical damping)
Steel lattice tower
- all welded
- welded bracings
- bolted bracings
0.2 %
0.2 - 1. %
0.3 - 1.2 %
Steel chimneys (unlined)
- welded
- bolted
0.4 - 0.7 %
0.6 - 1. %
Concrete chimneys
- without internal flues
- with internal flues
0.5 - 1.2 %
1.2 - 2.5 %
Steel tubes and pipes
0.5 - 1 %
Table 5.3: Indicative values of apparent fixity length for piles of oil
and gas jacket platforms [5.36]
Fixity length
Soil condition
3.5 D - 4.5 D
stiff clays
7D-8D
very soft silts
6D
general calculations
where:
D
outer pile diameter
5.2.2
Piled structures
The load transfer mechanism and consequently the modelling of piled foundations
differs between onshore and offshore wind energy converters.
On land a cross-shaped foundation plate of several hundred tonnes of concrete is
connected to 12 to 16 piles. So, the overturning moment and the base shear are
transferred into both axial and lateral pile loads. The piles are designed mainly for axial
capacity. For a dynamic analysis of the entire wind energy converter equivalent
foundation springs for the horizontal and rotational motion of the foundation plate are
introduced [5.41].
Piled foundation of offshore wind energy converters are either of the monopile or tripod
type. In the former case the foundation dynamics are entirely governed by the lateral
pile resistence while in the latter case also the axial resistence has a certain
contribution. A model of the actual pile or piles and of the surrounding soil along the
entire penetration depth is required. Attention has to be given to the properties of the
different soil layers, the inherent uncertainties of the estimated soil parameters and
scour degrading the foundation stiffness and strength.
91
Mtop , θtop
Mtop , θtop
dgen
dgen
Mtop , θtop
P
dgen
y
fixity length
ktrans
krot
(deq )
P
y
Node with degrees of freedom
Figure 5.7: Modelling of monopile support structures
depth below mudline [m]
Left:
Middle:
Right:
Simplified model with apparent fixity length
Simplified model with foundation springs at mudline
Detailed model of lateral pile behaviour by P-y curves
0
0
0
5
5
5
10
10
10
15
15
15
20
20
20
25
25
10 -400
25
-5
0
5
deflection [mm]
-200
0
resistance [kN/m]
200
0
10
20
30
bending moment [MNm]
Figure 5.8: Deflection, soil resistance and bending moment of a pile with 3.5 m
diameter in medium dense sand loaded by a thrust of 300 kN at the tower top
92
Chapter 5
Models of distinct sophistication are appropriate for the different stages of the design
process. Figure 5.7 illustrates such treatments which are demonstrated for the OptiOWECS monopile design (Figure 9.1).
During the preliminary design only crude estimates are possible. Most convenient is the
concept of the ‘apparent fixity length’, the length below the seabed under which the pile
is assumed to be rigidly clamped in order to achieve the same fundamental
eigenfrequency than the flexible supported pile (Figure 5.7 left). In the offshore
technology rough approximations are often used (Table 5.3). For large piles with
diameters of several metres probably lower values are more appropriate. For instance,
fixity lengths of 3.3 and 3.7 diameters are calculated by analysing one converter of the
first Dutch offshore wind farm Lely founded on sand and clay and the Opti-OWECS
design solution for medium dense sand, respectively (Section 9.3.1).
As an alternative, independent foundation springs, krot and ktrans, for the rotational and
translational flexibility could be used (Figure 5.7 middle). However, in distinction to the
gravity structures treated in the following section, it is difficult to find meaningful stiffness
values without assistance by a more sophisticated model. Randolph [5.42] has derived
analytical expressions for the foundation stiffness matrix, the deflection shape and the
distribution of pile bending moment by Finite Element analyses of soils with stiffness
proportional to the depth.
For both model types the material and soil damping and optionally also the aerodynamic
damping is included in the generalized damping of the different modes, dgen.
The conceptual and certainly the final design stages require proper modelling of the
subsoil behaviour (Figure 5.7 right). In accordance with common practice the following
simplified assumptions are made. The forces occurring between the soil and the pile at
any level are independent of the deflections at any other level. The effect of the mass
of the soil acting with the pile has a negligible effect on the structural dynamics.
The soil is represented by a finite number of lateral springs. The required grid spacing
can be related to the natural deflection wave length of the beam on an elastic foundation
[5.36]. For convenience one may select one metre between the springs for our
purposes.
Pile group effects do not need consideration if the distance between the piles is greater
than eight diameters. So, each pile of a tripod with three single piles can be modelled
analogously to a monopile except that also the axial deflection should be taken into
account.
The lateral load - deflection characteristics of the individual springs, denoted as P-y
curves, are derived by semi-empirical rules from the soil parameters and as a function
of the penetration depth. Although dedicated geotechnical codes are available for
complex conditions, the establishment involves always a certain degree of arbitrariness
and uncertainty. For more normal offshore structures in the petroleum industry the P-y
curves according to the API Recommendations [5.43] are quite popular and are
accepted as conservative estimates for the minimum bearing capacity. For an OWEC
both the lower and the upper bound of the soil stiffness are relevant and the API
Recommendations should be applied with care.
Figure 5.8 presents the lateral pile deflection, lateral soil resistance per unit length and
the pile bending moment for the Opti-OWECS monopile in medium dense sand.
93
resistance per unit length P [kN/m3]
10000
cyclic/static (25 m)
1000
static (5 m)
cyclic (5 m)
static (1 m)
100
cyclic (1 m)
10
0
10
20
30
40
50
deflection y [mm]
Figure 5.9: P-y curves for a pile with 3.5 m diameter and 25 m penetration into
medium dense sand
(Markers show operating points for 300 kN thrust loading at tower top.)
lateral load [kN]
600
load
amplitude
600 kN
400
load
amplitude
160 kN
200
0
-200
-4.0
0.0
4.0
8.0
12.0
deflection at mudline [mm]
Figure 5.10: Hysteresis loop of deflection at mudline for load cycles
with different magnitude
94
Chapter 5
The deflection of the pile remains relatively small. About 4 m below the sea bed the soil
shows maximum resistance. The maximum bending moment associated with the
loading by the rated aerodynamic thrust occurs 5.5 m below the soil surface. For pure
hydrodynamic loading with a much smaller effective lever arm of the applied load the
increase of the bending moment with depth along the first few metres is much steeper
and the critical cross section is located at a larger depth of up to 7 m.
Common offshore structures suffer very significant cyclic wave loads during extreme
conditions. Therefore API recommends different curves for static and cyclic loading
conditions. For an OWEC the case is more difficult since the fluctuating aerodynamic
and hydrodynamic loads are superimposed on the stationary mean wind loading.
Some of the P-y curves applied at the example are plotted in Figure 5.9. A difference
between the resistance for static or cyclic loading is observed only in the upper soil
layers, where however the maximum bending moment occurs. Markers on the curves
indicate the mean deflection associated with the rated thrust loading. Generally the
deflections for large monopiles are relatively small, i.e. in the range of millimetres. The
validity of the P-y curves according to API for such small strain magnitudes is more
qualitative. Therefore the differences between static and cyclic loading and whether a
pre-stress level is considered should not be overstressed. Other soil models can
generate quite different curves in the low strain region. Instead one should consider
such effects as part of a systematic variation of the structural uncertainties including
also the wide spreading of the soil parameters (Section 5.3).
Under the condition that the load variation does not alter significantly the soil stiffness,
the linearisation of the P-y curves for a certain pre-stress level is possible. This is
required for both frequency domain analysis and time domain simulation using modal
decomposition.
Soil damping of piled structures
Soil damping comprises two important parts: internal and radiation. Internal soil friction
resulting in hysteresis or plastic damping is more important for piles. Nonetheless, it is
often quite low for OWEC foundations with sufficient pile stiffness and penetration
depth. Under these conditions the load-displacement behaviour under operational
conditions is almost linear and therefore only small plastic deformations and associated
damping occur. Figure 5.10 shows the hysteresis loop of the displacement at the
mudline for the above mentioned monopile. The area enclosed by the loop corresponds
to the dissipated energy. The damping is negligible for a typical operational load
variation applied at a height of about 40% of the overall height. Only for extreme
conditions, e.g. an extreme operational gust with emergency shut-down, significant
damping occurs.
Closed form solutions for the hysteretic damping of piles exist only for completely plastic
and uniform soils [5.44]. Other conditions require a non-linear, static Finite Element
analysis of a load cycle with load amplitude and resulting deflection amplitude at
mudline, ŷ (Figure 5.10). The energy dissipation, Wdiss, during the load cycle is
represented by the area enclosed by the load path. An equivalent viscous damping at
mudline, deq (Ω), can be derived by equating the dissipated energy with the work of a
massless, spring-dashpot system during a corresponding load cycle with circular
frequency, Ω, (5.5).
deq ( Ω ) '
Wdiss
π Ω ŷ 2
(5.5)
95
Such a viscous damper can be included in the equations of motion but complicates the
modal decomposition. The damping is neither stiffness nor mass proportional and
results in either complex but uncoupled equations of motion or a non-diagonalised
generalised damping matrix and real, coupled equations. Here the latter is accepted
together with the assumption that the excitation frequency, Ω, equals the
eigenfrequency, 2 π fi , of the associated modal degrees of freedom, qi .
In Section 9.4 this exercise is performed for two offshore wind energy converters to
quantify the damping of different eigenmodes. In practice the low magnitude of the
hysteretic damping can be either neglected or it is considered directly as a small
addition to the material damping.
5.2.3
Gravity structures
The modelling of gravity support structures is similar to those of onshore wind energy
converters with a raft foundation. The caisson or foundation plate is considered as a
rigid body supported by a linear elastic, homogenous and isotrope half-space
characterised by the undrained density, ksoil , the dynamic shear modulus, Gdyn , and the
Poisson ratio, ν, (Figure 5.11 left). Finite Element models for the layered soil are
generally applied only for large offshore structures, nuclear power plants, high speed
railways or similar buildings.
The behaviour of the elastic half-space can be described by frequency dependent
springs and viscous dampers associated with the rigid body degrees of freedom of the
foundation (Figure 5.11 right). The springs are representing the elastic properties whilst
the damper elements model the radiation of vibration energy away from the caisson.
Material damping of the soil is generally neglected.
A variety of different and partially contradictory models are found in the literature to
handle the frequency dependency. Three categories can be identified.
• Frequency independent spring or spring and damper models (Figure 5.12 left)
Common agreement exists on the values of equivalent, frequency invariant foundation
springs as function of the soil parameters and the radius, R, of the foundation plate,
(5.6) [5.36, 5.45, 5.46]. For wind turbines on land normally only the tilt rotation and
horizontal translation are considered. Such models are widely used because of their
simplicity and the sensitivity study on the soil parameters which is carried out anyway.
For our purposes this fits to the preliminary and conceptual design stages.
8 Gdyn R
4 Gdyn R
8 Gdyn R 3
kvert '
ktilt '
khorz '
(5.6)
2 & ν
1 & ν
3 ( 1 & ν)
Frequency independent damping values are obtained by a dynamic analysis of the
entire system at resonance in the associated degree of freedom, (5.7, 5.8).
4.6 R 2
3.4 R 2
dhorz '
ksoil Gdyn
dvert '
ksoil Gdyn
(5.7)
2 & ν
1 & ν
Establishment of a frequency invariant parameter for the damping of the tilt rotation is
problematical since in principle a parabolic dependency on the excitation frequency and
no damping for the quasi-static case exist. Here we apply an approximative relation
including the inertia of the entire structure with respect to the centre of gravity, ΘtiltC.o.G..
96
Chapter 5
Mtop , θtop
dgen
k = f(Ω)
d = f(Ω)
Figure 5.11: Principle of modelling of gravity support structures
Mtop , θtop
Mtop , θtop
Mtop , θtop
dgen
dgen
dgen
k = k0
( d = d0 )
k(Ω)
+ i Ω d(Ω)
M1,trans
M1,rot
x1,rot
x1,trans
Node with degrees of freedom
Figure 5.12: Actual modelling of gravity support structures
Left:
Middle:
Right:
Frequency independent spring or spring and damper model
Frequency dependent spring and damper model
Frequency independent mass, spring and damper elements
with internal degrees of freedom, x1,trans and x1,rot
97
0.8 R 4 ksoil Gdyn
dtilt '
C.o.G.
(1 & ν )
1 %
3 ( 1 & ν ) Θtilt
(5.8)
8 ksoil R 5
• Frequency dependent spring and damper models (Figure 5.12 middle)
More adequate is the consideration of the frequency dependent properties. Values for
the stiffness and damping coefficient can be obtained from diagrams or formulas [5.46,
5.47]. Such models are suitable for harmonic analysis of structures in machine
dynamics, but they are difficult to use for transient analysis which is aimed for, here. In
addition, mode superposition techniques can hardly be applied [5.48].
• Simple physical models with frequency independent mass, spring and damper
elements (Figure 5.12 right)
More suited for the design practice are simple physical models with frequency invariant
mass, spring and dampers and with a few additional internal degrees of freedom [5.49].
This treatment approximates directly the transfer functions of the structure - soil system
derived by more complex models and can be included conveniently in the Finite Element
analysis of the structure. The approach is not further applied here because the nonproportional damping is not compatible with the standard mode superposition technique
and the damping effects are not so significant to justify the use of complex eigenmodes
to decouple the equations of motion.
In this thesis the frequency invariant spring and damper models according to the
recommendations of the German Geotechnical Society [5.45, 5.46] are applied for the
final design stage or detailed analyses. The relevant excitation frequencies are much
lower than in other applications, e.g. foundation of fast spinning turbines or foundations
with shock impact. In Section 9.4 the choice is justified by a comparison of different
models for a gravity monotower.
5.2.4 Direct aerodynamic loading of the support structure
During power production the magnitude of the aerodynamic rotor loads exceed the
direct wind loading of the support structure by an order of magnitude. Only beyond the
cut-off wind speed, when the turbine is parked or idling, the direct wind loading becomes
significant and has been considered in this thesis. The loading is applied as coherent
gusts based upon drag coefficients and quasi-static gust response factors in accordance
with DIN 4133 and DIBt Guidelines [5.50, 5.51]. Recently it became common practice
to simulate directly the dynamic loading by turbulent wind which leads to less
conservative results compared to the application of a gust response factor.2
5.2.5 Hydrodynamic loading
Classification of loading and analysis approach
Section 4.1.2 has discussed the first two steps of a hydrodynamic analysis dealing with
the determination of the environmental conditions and of the water particle kinematics,
2
In Section 4.1.1 it is mentioned that within this thesis the turbulent wind field is expressed in a rotating
reference frame. This has advantages for the rotor part, but makes it impossible to derive directly the wind
field affecting the fixed support structure.
98
Chapter 5
i.e. surface elevation, water particle velocities and accelerations. Here we address the
third step, the calculation of the hydrodynamic forces and response.
The fluid loading on structures may be classified as:
• Drag loading caused by vortices generated in the flow as it passes the structure. The
associated forces are proportional to the square of the relative particle velocity and
occur both in steady currents and wave fields.
• Inertia loading generated by the pressure gradient in an accelerating fluid moving
around a member. The inertia force is proportional to the acceleration of the water
particles owing to waves.
• Diffraction loading, a part of the inertia force on a structure which is so large that the
wave kinematics are altered significantly by its presence.
• Water added mass, an inertia force experienced by a member accelerated relative
to the fluid.
• Slam and slap loading. An inertia force proportional to the square of the relative
velocity when a member passes though the water surface. Here it is mentioned as
one approach for breaking wave loading (Section 4.1.2).
• Vortex shedding induced oscillating loading associated with the drag loading. Forces
with the frequency of vortex separation are generated lateral to the flow direction and
with the double frequency longitudinal to it. For our purposes with relatively large pile
diameters it is without relevance.
The wave force approach to be applied is determined by two non-dimensional
quantities. The non-dimensional Keulegan-Carpenter number, KC, increases with the
ratio between the drag and inertia force, FD / FM. The relative size of the structure,
expressed by the diameter of the structural member, D, divided by the wave length, L,
determines the degree of wave diffraction.
Figure 5.13 locates two common methods in relation to the Keulegan-Carpenter number
and relative size of the structure. The diagram is limited on both axes by the slope of the
deep water breaking wave curve, H/L = 1/7, where H refers to the wave height.
The Morison equation obtains the total force from an inertia and a drag component.
Hydrodynamic drag becomes significant when the diameter of the structure is small
compared to the wave length, D/L < 0.2 and the wave height, H, is large with respect to
the water depth (upper left corner of Figure 5.13).
For deep water the Keulegan-Carpenter number is proportional to the ratio of wave
height and structural diameter, (5.9), whilst for lower water depth it is driven by the
amplitude of the water particle velocity.
π H
for water depth d > L / 2
KC '
(5.9)
D
Taking the example of a monopile with 3 m diameter and waves up to 1 m height, typical
for fatigue conditions, KC is equal or smaller than 1, corresponding to a drag force
below five percent of the inertia force (Figure 5.14). The drag force should however not
be underestimated for extreme conditions. High waves and near-breaking waves in
shallow water are associated with high velocities in the wave crest and the non-linear
drag force becomes of similar magnitude than the inertia loading, even for compact
monopiles (Figure 5.15).
99
Figure 5.13: Ranges of application of wave force formulas [5.36]
If the structure is large in relation to the wave length, for example for gravity structures,
the effects of diffraction and scattering have to be considered by the diffraction theory.
It is based on a potential flow theory and requires in most cases a numerical solution of
the Laplace equation and the boundary conditions [5.52].
For our purposes, dealing mainly with monopile and tripod configurations, the Morison
equation is more relevant and diffraction can be accounted for by suitable corrections.
Implementation in a design tool, as for instance the DUWECS code, is straightforward.
Morison equation
According to Morison et al. [5.53] the hydrodynamic force per unit length, dF, for a fluid
with density, kw, is represented by an inertia term, which is linear in the water particle
accelerations, üw, and quadratic in the structural diameter, D, and a drag component,
that is quadratic in the water particle velocities, u0 w , and linear in D, (5.10).3 The
expression is empirical owing to the drag and inertia coefficients, CM and CD , which are
determined from experiments.
3
In accordance to the notation in structural dynamics we denote the displacement, velocity and
acceleration as u, u0 and ü. In contrast, u and u0 are commonly used for the velocity and acceleration in
hydrodynamics.
100
Chapter 5
dF '
CM kw
π D2
ü w
4
inertia force on
fixed structure
%
CD kw
D
2
u0 w u0 w
(5.10)
drag force on
fixed structure
The equation is valid for most bottom-mounted OWEC support structures loaded by
higher waves. Exceptions are the caissons of gravity support structures and tower
structures with a large diameter in small waves when diffraction becomes important. In
both cases drag is negligible but diffraction has to be accounted for.
Initially the Morison equation was proposed only for the horizontal wave force on a
vertical cylinder. It shown, however, also to give reasonable results for the more general
case of inclined cylinders when applied to the flow and force components resolved
normal to the member axis.
The drag force, proportional to the square of the water particle velocity, is non-linear
even if the water particle velocity is sinusoidal according to the linear wave theory. This
generates difficulties in linear frequency domain and probabilistic analyses. From a
dynamic viewpoint the non-linearity is important since a part of the wave energy is
transferred to the region of the structural eigenfrequencies. The amplitude of the third
and fifth harmonics of the wave frequency account for 1/5 and 1/45 of the drag force.
For illustration the contribution of the drag and inertia loading to the total base shear
was calculated for the monopile of the Utgrunden wind farm with 3 m diameter in 10 m
water depth. In the fatigue regime with moderate but frequent waves (Figure 5.14) the
inertia forces dominate and the behaviour is linear. In contrast, both the water
kinematics and the loading becomes non-linear at extreme conditions. Figure 5.15
shows kinematics and forces associated with a wave of 6.4 m height. The wave crest
of 4.95 m is more than three times higher than the trough which results in high particle
velocities in the crest, resulting in a drag force of almost the same order of magnitude
as the inertia force. The bending moment due to the wave impact (not shown) contains
several higher harmonics with significant excitation energy.
Structural motion, hydrodynamic added mass and damping
The Morison equation can be applied also to a structure moving in a calm fluid or in a
wave and current field with structural acceleration, ü, and velocity, u0 , when it is
formulated in the relative water particle kinematics.
πD 2
π D2
D
ü w & ( CM & 1 ) kw
ü % CD kw
u0 w & u0 u0 w & u0
dF ' CM kw
4
4
2
(5.11)
inertia force on
water added
drag force incl.
fixed structure
mass force
hydrodyn. damping
For numerical purposes it is convenient to split up the expression of the inertia force in
terms of the water particle acceleration and structural acceleration, (5.11). Successively,
the product of the so-called 'water added mass' and the acceleration of the cylinder can
be written to the left hand side of the equations of motion. The dry structural mass of the
submerged elements is increased by this portion which equals approximately the
amount of gross displaced water since the value of CM is around two. The additional
mass is considered only up to the constant still water line in order to keep the mass
matrix time invariant.
(Utgrunden monopile ø 3 m, water depth d = 10 m))
Figure 5.14: Kinematics and forces of an inertia dominated
fatigue wave calculated with linear wave theory
(Utgrunden monopile ø 3 m, water depth d = 10 m)
Figure 5.15: Non-linear kinematics and forces
of an extreme wave in shallow water calculated with
10th order stream function wave theory
101
102
Chapter 5
validity of
Morison equation
diffraction theory or
correction of Morison equation required
inertia coefficient C_m [-]
2.25
2.00
1.75
1.50
1.25
1.00
0.75
0.50
0.25
0.00
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
cylinder diameter / deep water wave length (D/L) [-]
Figure 5.16: Effect of wave diffraction on the inertia coefficient CM of a
circular vertical cylinder
Table 5.4: Chosen hydrodynamic coefficients
Parameter
†
Fatigue conditions
Extreme conditions
inertia coefficient CM
2.0
1.7
drag coefficient CD
0.7
0.6 †
0.7 ‡
smooth without marine growth, ‡ rough with marine growth
The drag force on the moving structure generates hydrodynamic damping. This viscous
force depends on the square of the relative particle velocity. In a time domain analysis
this is handled by an iteration procedure which is started with the values of the
structural velocity from the last time step. In the frequency domain the wave drag force
and the hydrodynamic damping are linearised.
As mentioned-above, Section 9.4.2 proves this effect only theoretically relevant for
bottom-mounted OWEC since the water depth is low and the structures are relatively
stiff in the submerged part which results in negligible structural velocities in this region.
Selection of hydrodynamic coefficients and diffraction correction
The hydrodynamic coefficients CM and CD are obtained empirically. They depend mainly
on the Reynolds number, the Keulegan-Carpenter number KC, the shape of the cross
section and the specific roughness due to marine growth. Moreover the flow situation,
e.g. oscillating fluid or structure in waves, has some influence. Establishment of
appropriate values requires experience and/or laboratory testing.
In design practice often fixed values, not corresponding directly to these complex
influences are chosen. They are valid only in a certain context as they are compensating
103
other, conservative assumptions. Table 5.4 lists the recommendations of the UK
Department of Energy [5.37], which are followed here under the conditions that
independent (extreme) values of wave and current are combined and that unidirectional
(long-crested) waves, regular wave theories and no shielding effects are considered.
For hydrodynamically compact designs the disturbance of the wave field by the structure
and the reflection of wave energy becomes important and requires specific
computational tools based on solution of the potential flow problem. For a single column
structure the Morison equation may still be applied in the diffraction regime by
consideration of a modified, frequency dependent inertia coefficient and a phase shift
of the wave kinematics according to MacCamy and Fuchs [5.54] (Figure 5.16).
A further documentation of the hydrodynamic models applied in this thesis can be found
in reference [5.55].
5.2.6 Sea ice loading
Prediction of static or dynamic ice loads is difficult due to physical, statistical and model
uncertainties. Beside the spreading of the environmental properties there is a variety of
empirical approaches. Common ice pressure models include up to four empirical
factors.
The actual ice loading is caused by current and wind induced ice motion and the failure
of the ice floe at the structure. At low failure frequencies of up to 1 Hz the ice sheets are
broken by the bending mode and an ice cone can reduce the loading very effectively.
For higher frequency of 0.5 to 10 Hz the ice is crushing and dynamic ice-structure
interaction occurs.
Figure 5.17 illustrated four ice failure modes distinguished by increasing ice velocity.
Examples for static or quasi-static load equations appropriate for mode I and II are
provides by references [5.57 - 5.59]. At modes III and partly mode II a steady state
interlocking of the ice failure frequency and the structural frequencies can occur and
very significant dynamic amplification factors are the result if the structure is poorly
damped. Some researchers propose a sinusoidal force history [5.59, 5.56] while other
u
u
t
Mode I: almost steady ice
t
Mode II: periodic transient response
F
u
t
u
t
Mode III: steady vibration
due to interlocking
t
Mode IV: small response due to
continuous ice crushing
Figure 5.17: Structural displacement u as response on different ice failure modes
distinguished by increasing ice velocities [5.56]
104
Chapter 5
recommend a saw-tooth like function which includes higher frequencies as well [5.60].
5.3
Particular sub-system aspects that affect the entire system
For the understanding of the dynamic behaviour of an OWEC it is essential to mention
some particular sub-system aspects that have a major effect on the overall dynamics.
Time invariance of the rotor in an inertial reference frame
The variation of the rotor stiffness and inertia in the inertial reference frame introduces
some difficulties in the treatment of the dynamics. The effect, well known for wind
energy converters, is of particular importance for (one and) two bladed rotors and to a
lesser extent for flexible three-bladed rotors in general. Different established methods
are available within the time domain approach. In the frequency domain the treatment
of a linear model with periodic coefficients is possible as well but not yet used for design
purposes [5.61]. For modal analysis, frequently a description a rotational invariant,
isotopic rotor or a fictitiously fixed or ‘frozen’ azimuth position is applied. The latter is
done within this thesis and all mentioned eigenfrequencies are so-called ‘snap-shot’
natural frequencies.
Aerodynamic damping
Any oscillating structure located in the air, for instance a chimney, experiences some
amount of damping which is generally lower than the material damping. For wind
turbines a different effect is of paramount importance, i.e. the aerodynamic damping
induced by the operating rotor.
Figure 5.18 explains the effect for one oscillation cycle of the structure. A motion of the
tower top in the wind direction (second figure) results in a smaller angle of attack, α, at
the rotor blades because the apparent out-of-plane velocity component, V (1-a), is
reduced by the tower top velocity, u& top . A smaller angle of attack corresponds, for
attached flow conditions, to lower aerodynamic lift and drag forces and to a reduction
of the thrust force, dFx, by an increment, ∆dFx. Likewise a movement of the nacelle
against the wind direction increases both the angle of attack and the thrust force. In both
situations the alternation of the thrust is oriented opposite to the disturbing tower top
motion and is experienced as aerodynamic damping. Similar but less pronounced
interaction occurs for the other tower top degrees of freedom.
For higher angles of attack, stall occurs and the aerodynamic damping is lower or may
even become negative because the slope of the lift curve is reduced. Under such
conditions aero-elastic instability can occur if insufficient damping exists.
Equation (5.12) provides an instructive, analytical expression for the damping ratio as
fraction of critical damping. The relation is based on stationary rotor aerodynamics and
some simplifications valid for a wind turbine operating with a high tip speed ratio and
near the rated wind speed.4 A similar form including the effect of the aerodynamic drag
can be derived and is more appropriate for stall regulated wind turbines above the rated
wind speed.
4
Garrad has described the background of the formula in reference [5.62].
R
ξaero '
Nb k Ω
/0 c ( r ) r dr
d α 00 r
d cl
8 π f o Mo
105
(5.12)
Rroot
where:
ξaero
Nb
kair
Ω
fo
Mo
Rroot , R
dcL / dα
c(r)
r
aerodynamic damping ratio as fraction of critical damping
number of blades
air density
rotational speed of the rotor
[rad/s]
frequency of the first fore-aft mode
[Hz]
modal mass of the single degree of freedom system
blade root rotor radius, outer rotor radius
slope of the lift coefficient with respect to angle of attack
chord length at radius r
spanwise coordinate
The closed-form estimate discloses two main influences on the damping. Firstly the
slope of the lift coefficient versus angle of attack can vary from approximately 2B for
attached flow to zero or even negative values in case of separated flow. An effect that
cannot be covered by the simple model is the critical dependency of the aerodynamic
damping in separated flow on non-linear unsteady stall hysteresis effects where it is not
simply related to the slope of the lift coefficient. Secondly, the aerodynamic damping
ratio depends on the eigenfrequency and modal mass of the support structure. It can
change from less than 1% of critical damping for a stiff and heavy tower to above 5%,
several times the structural damping, in case of a soft and light support structure.
Considerable aerodynamic damping takes place only during the operation of the rotor
and is active on both aerodynamic and hydrodynamic excitation of the system. Apart
from the fore-aft motion of the support structure motion also the flapwise, i.e. out-ofplane, motion of the flexible blades and to a lesser extent for the tilt and yaw movement
of the nacelle experience aerodynamic damping.
Any dynamic OWEC model must account for aerodynamic damping in a reasonable
way. Integrated, non-linear time domain simulation are most accurate (Chapter 6).
During the early design stages simplified models are advantageous (Chapter 8) which
substitute the aerodynamic damping by additional structural damping.
Variation of system properties
Most of the system properties discussed in the last three sections can only be estimated
with a considerable uncertainly; moreover, they are to a certain extent time variant due
to effects of operating conditions, loading, tide, marine growth, aging, settlement, etc.
Therefore either design calculations have to account for these variations or should
assume at least the most onerous combination(s). Tables 5.5 and 5.6 list conditions
resulting in either maximum or minimum values of the eigenfrequencies and damping.
A priori it is not always clear for which conditions the highest extreme response will
occur. For fatigue loads which accumulate over a long time it might be appropriate to
assume average conditions, e.g. mean sea level, moderate marine growth, low lateral
pile loading, etc. So different structural models of the OWEC should be established for
fatigue and extreme load calculations assuming the most onerous conditions.
106
Chapter 5
A
A
x&
x&
top
=0
x&
Fx
top
Fx
V(1 - a) - x&
✕r
V(1 - a)
V(1 - a)+ x&
top
cL
top
cL
cL
α
✕r
α
α
✕r
α
A-A
α
α
A-A
dFx
dFx
dFx
dFx
dFx
α
α
dLift
α
v
top
dDrag
Figure 5.18: Illustration of aerodynamic damping effect
• fore-aft motion u& top at wind speed V during production
• motion u& top changes angle of attack α at a blade element
• variation of lift coefficient cL and drag cd (not shown)
• variation of thrust ∆Fx opposite to perturbation
107
Table 5.5: Variation of system properties for calculation of eigenfrequencies
Parameter
mass effects
still water level (SWL)
hydrodyn. added mass
internal fluid
marine growth
soil mass
structural mass
rotor position (2-bladed) †
wind turbine mass /
inertia
Reduction of
eigenfrequency
Increase of
eigenfrequency
HAT + surge + settlement
max CM
associated with max SWL
max
assuming max density
max
vertical
yes
LAT - surge
min CM
associated with min SWL
none
assuming min density
min
horizontal
min or none ‡
stiffness effects
soil stiffness
min, P-y curves for static
max, P-y curves for cyclic
loading
loading
max
none
mean lateral pile loading
max
none
scour
†
resulting in minimum or maximum eigenfrequency of the fore-aft bending modes
‡
during installation and major overhaul condition
Table 5.6: Variation of system properties and operational conditions for calculation
of damping
Damping type
Minimum damping
Maximum damping
aerodynamic damping
parked or idle rotor or
operating rotor with
separated flow,
lower bound of Ω,
maximum bending
eigenfrequency
operating rotor with
attached flow,
upper bound of Ω,
minimum bending
eigenfrequency
radiation damping of
gravity foundation
lower bound of soil shear
strength and density, low
excitation frequency
upper bound of soil shear
strength and density,
high excitation frequency
hysteretic damping of pile
foundation
upper bound of static soil
stiffness, no mean lateral
pile loading
lower bound of cyclic soil
stiffness, upper bound of
mean lateral pile loading
structural damping
lower bound
upper bound (e.g.
cracked concrete)
hydrodynamic damping
(if relevant)
none or related to min CD,
min still water level, low
waves
related to max CD, max
still water level, high
waves
108
Chapter 5
109
Integrated, Modular, Non-Linear Time Domain Simulation
CHAPTER 6
INTEGRATED, MODULAR, NON-LINEAR TIME DOMAIN SIMULATION
Non-linear time domain simulation of the overall dynamics of an entire offshore wind
energy converter is a very powerful and potentially very accurate analysis method. This
chapter describes both the general principles and the particular implementation in this
thesis, denoted here as integrated, modular, non-linear time domain simulation.
In order to achieve its full potential, however, sufficient experience and high
computational effort in terms of dedicated mathematical models and computational time
is required. Therefore in engineering practise the approach is best suited for in-depth
investigations and certification calculations at the end of the design process when
detailed design data are available and a good understanding of the principles underlying
the design has already been gained.
During parameter studies and in the early stage of the design process a simplistic
application with an inadequate structural model or random simulations which are not
representative for the actual stochastic loading can be dangerous. The accuracy might
then be lower than that of simplified and faster approaches as introduced in Chapter 8.
Wind Turbine
Controller
Wind Field
Waves,
Current, Ice
Electro-Mechanical
Conversion System
Aerodynamic
Loads
Rotor
Hydrodynamic
Loads
Support
Structure
Grid
(Tower &
Foundation)
Soil
Soil Dynamics
Offshore Wind Energy Converter (OWEC)
Environment
Dynamic Influence:
major
Consumption
minor
Figure 6.1: Integrated dynamic model including short-term dynamic interactions
within offshore wind energy converter (OWEC) and between OWEC and both natural
environment and grid
110
Chapter 6
The time domain simulation approach will be elaborated with respect to a number of
general aspects of structural dynamics and practical requirements.
Section 6.1 explains the importance of an integrated model of an entire offshore wind
energy converter and how it is implemented in a modular manner in this research. So,
the convenient combination of distinct, but dedicated structural models in the wind
turbine and support structure part is possible.
The decision between time domain and frequency domain approach is discussed in
Section 6.2.
Section 6.3 investigates techniques for the reduction of degrees of freedom and of load
cases with long duration of the simulations, while Section 6.4 provides an overview on
the implementation in the DUWECS (Delft University Wind Energy Converter Simulation)
code.
6.1. Integrated but modular dynamic model
Integrated model
6.1.1
Physical background
A simple model of an offshore wind energy converter considers a system with two
independent, random vectorial inputs for the aerodynamic and hydrodynamic excitation.
Figure 6.1 illustrates a more detailed block model showing the simultaneous, shortterm interactions between the aerodynamic, hydrodynamic and geotechnical
environment, the offshore wind energy converter with its major components and the
consumption of electrical energy.
+
x&& +
+
x& +
+
x =
+
[ M WT + SS + M hSS ] x&& + [ D WT + SS + D aWT + SS ] x& + [ K WT + SS + K aWT + SS ] x = f aWT + SS + f hSS
x WT
x=
x SS
Figure 6.2: Structure of the equations of motion of an integrated OWEC model
WT
: wind turbine
M
: mass matrix
SS
D
: damping matrix
: support structure
K
: stiffness matrix
: aerodynamic
a
: hydrodynamic
x
: degrees of freedom
h
f
: load vector
Occupied portions of the matrices are highlighted. The wind turbine related
portions of the structural mass, damping and stiffness matrices are generally time
variant and depend upon the rotor speed, azimuth angle and deformation of both
rotor and structure. The aerodynamic damping and stiffness matrices depend
also on the wind speed. The hydrodynamic mass matrix is constant.
111
Although the random aerodynamic and hydrodynamic loading can be regarded as
independent under fatigue conditions it is not straightforward to calculate the
aerodynamic and hydrodynamic response separately. We will discuss this further in
Section 7.1. Considering the constitution of the linearised equations of motion in Figure
6.2, we find aerodynamic damping and stiffness terms and a hydrodynamic inertia term
in the system matrices. So, the responses to any of the two independent load inputs will
depend on both aerodynamic and hydrodynamic conditions.
Design practice
For some years, integrated dynamic models in the time domain are the only accepted
approach for design calculations of onshore wind energy converters. Section 5.1
explained how such models integrate aerodynamics, structural dynamics of rotor, drive
train and tower as well as the dynamics of the generator and control system. From the
discussion above it is consequent that offshore wind energy converters are modelled in
an integral way by considering the wind turbine, with all above-mentioned features, and
the support structure, their simultaneous response under aerodynamic, hydrodynamic
and electrical loading and their mutual interactions (Figure 6.1). This is denoted as an
integrated model within this thesis.
However, is an integrated model really required and does it fit the needs of an iterative
design process?
A separate treatment of the main sub-systems, the wind turbine and the support
structure, would be convenient with respect to the complexity of the design problem and
the expertise distributed on the two disciplines of wind energy and offshore technology.
There is no single answer to this question for all applications. From a scientific point of
view the integrated model is the obvious choice and it should be used for any
investigation of OWEC dynamics.
During the design process, resources are scarce and must be used efficiently. Evidence
exists that the wind turbine part is not significantly affected by wave induced response.
The support structure, however, is suffering simultaneous wind and wave response and
incorporation of the aero-elastic interactions, namely the aerodynamic damping, is
essential [6.1, 6.2].1
For analysis limited to the wind turbine part of the OWEC, application of an onshore wind
turbine design tool might be sufficient if the dynamics of the tower and foundation are
properly represented. Furthermore, it has to be ensured that such a reduced model is
consistent with the wind turbine representation within the integrated OWEC model used
for the support structure analysis.
During the early design stages or if fatigue is not governing the design, the support
structure fatigue might be analysed with one of the simplified methods based upon
separated wind and wave responses (Section 8.1 and 8.2). The integrated approach is
clearly required for certification calculations of structures dominated by combined wind
and wave fatigue.
1
Section 9.3.2 explains why the wind turbine rotor is insensitive to the wave excitation. The extent of
interactions and the conditions under which separation of the simultaneous response is possible for
simplified analyses is investigated in Section 7.1.
112
Chapter 6
Table 6.1: Comparison of wind turbine and support structure
with respect to modelling requirements
Aspect
Wind turbine
(horizontal axis)
Support structure
(bottom-mounted)
Geometrical
configuration
(flexible) multi-body
system with open chains,
limited set of generic
types (e.g. 2- or 3-bladed,
fixed or flexible yaw, etc.)
fixed structure of
cantilever or truss type,
various configurations
especially for braced and
lattice type
Structural kinematics
flexible deflections
superimposed on gross
motion (e.g. azimuthal
rotation, teeter, tower top)
(small) flexible deflection
about a mean position
Stress-strain material
behaviour
linear
linear, partly non-linear for
piled foundations and/or
extreme conditions
Material damping
proportional
proportional
Other damping
aerodynamic damping
aerodynamic damping,
equivalent viscous soil
damping
Structure-flow
interaction
important
less important
Equations of motion
non-linear, time-variant
linear, time-invariant
Typical number of
degrees of freedom
6 - 30 (modal plus
discrete) †
20 - 600 (physical),
5 - 30 (modal) ‡
aerodynamics, gravity,
manoeuvre
mainly non-linear
- random
- deterministic
wind turbine, gravity,
hydrodynamics, ice
linear†† or non-linear‡‡
- random
- deterministic or random
Dynamic loading
origin
nature
- fatigue conditions
- extreme conditions
†
‡
††
‡‡
valid for modal representation of the wind turbine (Section 5.1),
valid for Finite Element representation of the support structure (Section 5.2),
fatigue conditions and/or inertia dominated structures,
extreme conditions and/or slender, drag dominated member
113
In any case an integrated OWEC model is an important tool that can be used in three
ways:
• Investigation of the dynamics of the entire OWEC and of the support structure
response to simultaneous aerodynamic and hydrodynamic loading during the final
design stages when simplified methods are not applicable.
• Studying the wind turbine response to pure wind excitation. For convenience all
hydrodynamic features are switched off and the integrated model behaves similarly
to an onshore design tool.
• Investigation of the hydrodynamic response of the support structure for an extreme
load analysis and parameter studies. Now the model acts as an offshore technology
design tool.
If properly applied, an integrated model does not only provide a more accurate
description, it also enables less conservative and thus most probably more cost-efficient
design solutions. This might explains the growing commercial interest in the approach.
Overview of integrated design tools
Today, several integrated OWEC design tools in the time domain exist. The early model
of Oscar and Paez based upon the general purpose Finite Element code NASTRAN™
[6.3] and the extension of an offshore technology design tool by Kortlever [6.4] are no
longer in use. All other programs are extensions of onshore wind turbine design tools.
Continuation of the work of Wastling and Quarton [6.1] led to the first commercial OWEC
design tool Bladed for Windows [6.5]. Approximately at the same time the DUWECS
code was further developed within the work summarised in this thesis and was applied
during different research and industrial projects at Delft University of Technology.
Quite recently, other commercial onshore wind turbine design tools as PHATAS, FLEX5
and HawC have been extended [6.6 - 6.8], but only illustrative results have been
published so far.
Non-linear, periodic and unsteady effects in the wind turbine part (Section 6.2.1) and the
interaction between wind turbine and support structure complicate the derivation of
frequency domain models. Such tools have been proposed, e.g. [6.1, 6.6, 6.9], but these
linear models are not fully integrated because spectral superposition of separate wind
and wave responses is assumed. Their application is limited to research or the early
stage of the design process.
6.1.2
Modular model
Distinct modelling requirements
The wind turbine and support structure sub-systems are subject to quite different
modelling requirements (Table 6.1). For the wind turbine, only very few geometrical
configurations are relevant which all result in geometrically non-linear, time-variant
equations of motion owing to the gross rotation and flexible deformation of the rotor. The
support structure properties are just opposite. Figure 3.16 shows a variety of structural
configurations, e.g. monopile, tripod, lattice tower. Nonetheless, all equations of motion
are linear and time-invariant. Likewise, the non-linearity of the external forces is less
pronounced for the majority of the hydrodynamic load cases than for the aerodynamic
conditions.
114
Chapter 6
There are two principal approaches to integrate such different requirements in one entire
model:
• consistent formulation of the equations of motion for the entire system
Typically this could be done by a sophisticated Finite Element representation capable
of gross motion superimposed on small deflections or by a multi-body system code
with automated generation of the equations of motion. This option offers a more
consistent description, but requires extensive computational power.
A number of recently extended onshore wind turbine tools provide a compromise in
this respect. They are always dealing with the same generic set of equations of
motion by consideration of only one support structure type, namely the simple
monopile/monotower configuration. Sometimes other structures can be modelled by
importing externally generated mass and stiffness matrices.
• modular formulation of the equations of motion for the wind turbine and support
structure
Obviously a multi-body or modal approach with specific treatment of the aero-elastics
fits the wind turbine part whilst a linear Finite Element model is ideally suited for the
support structure. For both aspects established models exist in the form of onshore
wind turbine design tools and multi-purpose structural codes. The approach is more
elegant and efficient, but requires a solution of the interface problem between the two
sub-systems and associated models. The greater computational efficiency during the
actual analyses is of particular value for time domain Monte-Carlo simulations.
Within this research we followed the second option and extended the (onshore) wind
turbine design tool DUWECS (Section 6.4) to offshore wind energy application. Among
other reasons this code was chosen due to its suitability for the modular approach.
The equations of motion of the used rotor module [6.10] are derived with the so-called
‘method of generalised momenta’ after van Holten [6.11]. This avoids a cumbersome
second time derivative known from Lagrange equations, the principle of virtual
displacements or most other approaches. Previous work by the author adopted the
method of generalized momenta to the modular approach within DUWECS [6.12].
Coupling of equations of motion of wind turbine and support structure
Several established methods exist for the treatment of the crucial interface between the
rotating and non-rotating structure [6.13 - 6.15]. The DUWECS code applies a different
approach, the stepwise integration of the equations of motion of the separated modules,
which is possible only in the time domain.2
The complete system is divided into a number of modules under the assumption that the
interaction between the modules takes place by prescribed variables (Figure 5.6).
Typically such interaction variables are either kinetic variables, e.g. generalised forces
and momenta, or kinematic quantities, e.g. generalised coordinates and velocities
(Figure 6.3). The equations of motions are established separately for every module
taking into account global inputs and outputs, e.g. wind speed or grid voltage, and the
2
This treatment has been applied also in other technical fields, e.g. the contact dynamics of the
current collector - overhead line system in railway engineering [6.16]
115
kinetic interaction variables
Fxx , Fyy , Fzz , Mxx , Myy , Mzz
Tower
Rotor
kinematic interaction variables
duxx /dt
or duxx /dt , duyy /dt
Figure 6.3: Rotor - tower interaction in the used DUWECS version
interaction variables from the other modules. This feature provides a large degree of
freedom to the code developer as long as the module interfaces are compatible. In
particular it enables use of the above-mentioned combination of a non-linear multi-body
system approach for the rotor and of linear Finite Element methods for the support
structure.
Instead of assembling the coupled set of equations of motion of the entire system, the
equations of the modules are integrated separately and subsequently for every module.
Depending of the module sequence, it is required to approximate the instantaneous
value of some interaction variable by the value of the previous integration time step. For
instance, the rotor behaviour is calculated based upon the tower top velocities from the
last time step while the tower motion is considering the tower top forces from the current
time step, which were calculated just before in the rotor module.
actually used
degrees of freedom
Ux
error with respect to application of all 6 degrees of freedom
Ux, Uy
U x , U y , Rot x
U x , U y , Rot y
U x , U y , Rot z
U x , U y , Rot x , Rot y
U x , U y , Rot y , Rot z
U x , U y , Rot x , Rot z
-15.0%
-12.5%
-10.0%
1st fore-aft mode at 16 m/s (stall)
-7.5%
-5.0%
-2.5%
0.0%
2.5%
error [%]
1st fore-aft mode at 10 m/s (attached flow)
Figure 6.4: Error of aerodynamic damping of first fore-aft mode by using only certain
tower top degrees of freedom at the example of a stall regulated wind turbine
(x: in wind direction, y: lateral to the left, z: upwards, see Figure 0.2)
116
Chapter 6
Experience shows that for the typical length of the integration time step between 5 and
25 ms no iteration of the interaction variables is required to simulate the overall
behaviour in a relevant frequency range of up to approximately 5 Hz.
The incorporation of the generalised tower top forces in the dynamics of the support
structure is straightforward; however, consideration of the tower top motion involves a
lot of bookkeeping in the derivation of the equations of motion in the rotor module.
Therefore several wind turbine design tools currently applied by the industry suffer
shortcomings in modelling the tower top motion.
The rotor module of the DUWECS code generates the entire set of six generalised tower
top forces but, depending on the used rotor module, either only the kinematics of the
fore-aft translation or only the fore-aft and lateral translation are considered in the
dynamics (Figure 6.3).
In a simplified manner we attempt to judge the error in the aerodynamic damping of the
first fore-aft bending mode which results from the incomplete modelling of the rotor tower interaction. From Kaiser [6.17] and Peinelt [6.18] we take the aerodynamic
damping matrix for the entire set of six degrees of freedom at the tower top of a stall
controlled 270 kW wind turbine. Such a damping matrix can easily be incorporated into
a general purpose Finite Element model of the wind turbine - support structure system.
By switching on and off different degrees of freedom we can now observe the
consequences for the damping of the fore-aft mode.
For this particular case, the error is conservative and in an acceptable range between
6% and 15% if only the tower top fore-aft translation, ux , or both fore-aft and lateral
translation, ux and uy , are considered (Figure 6.4). Activating the tilt degree of freedom,
roty, and especially of both the rolling and tilt degrees of freedom, rotx and roty, further
reduces the error with respect to the complete modelling.
The simplified modelling will likely cause larger errors in the description of the
aerodynamic damping of the second fore-aft mode which is however less important for
the hydrodynamic response.
6.2. Time domain versus frequency domain approach
6.2.1
Non-linearities
The three different media, air, water and soil, all behave non-linearly with respect to the
loading and response of the offshore wind energy converter.
The strongest non-linearity occurs in the rotor aerodynamics due to flow separation at
higher wind speeds. The stall effect and the associated variation in the slope of the lift
coefficient discard linearisation of the aerodynamics (Figure 5.5).
Kinematic non-linearities of the hydrodynamics are significant for high waves and in
shallow waters. The finite wave height and the asymmetry of crest and trough are
incompatible with linear wave theory and lead to higher harmonics in the water particle
velocities and accelerations.
For our purposes the non-linearity of the hydrodynamic drag is of lesser importance,
since most OWEC support structures are dominated by the hydrodynamic inertia force.
Under such conditions the offshore experience with the linearisation of the drag force
is applicable and linear spectral analysis is possible. The drag becomes, however,
117
significant for extreme, (near-)breaking waves as OWEC structures are located in small
water depths. The hydrodynamic non-linearities can result in excitation of structural
frequencies by higher order harmonics of the wave frequency, the so-called ringing
effect [6.19].
In principle, the pile-soil interaction under static or cyclic loading is non-linear and the
soil behaviour of a gravity foundation depends on the excitation frequency. In Section
9.4.3 such effects are judged to be less significant.
As a reminder to the discussion of Section 6.1.2, the geometrical non-linearities caused
by the gross overall motion of the rotor and the considerable elastic deflections of the
slender blades are mentioned again.
The long term fatigue behaviour of the structure establishes another strong non-linearity,
which is however independent of the structural model. The fatigue deterioration is a
function of the third or higher power of the stress range. In contrast with the material of
the structure itself can be treated according to classical theory of linear elasticity.
To summarize, significant non-linearities and transient effects are present, especially in
the wind turbine part, which explains the preference of the wind energy community for
a non-linear analysis approach for this sub-system. Linear treatment of the support
structure is reasonable for most fatigue load cases, but is limited for extreme conditions
in shallow waters.
Fatigue analysis in time domain or frequency domain
6.2.2
Different methods are established for each of the four steps included in a fatigue
analysis, i.e. stochastic environmental modelling, structural response calculation,
establishment of stress range distribution and damage accumulation. There are only two
basic approaches commonly used (Figure 6.5):
• time domain approach, i.e. generation of random time series from wind and wave
spectra, non-linear, unsteady system simulation by time step integration techniques
and Rainflow counting of stress ranges
• frequency domain approach, i.e. direct application of the environmental spectra, linear
spectral analysis and determination of stress range distribution from the spectral
moments
The applied damage accumulation hypothesis is largely independent from time or
frequency domain approach. Both wind energy and offshore technology commonly use
the well-known linear damage accumulation rule after Palmgren-Miner together with
standard S-N curves. More sophisticated approaches based upon fracture mechanics
are sometimes applied in offshore engineering [6.20].
Indicated by dashed lines, Figure 6.5 shows some alternative paths, which are
uncommon due to the required effort or the lack of certain information.
Time domain approach
The rationale of the time domain approach is to simulate the physical behaviour
of the system by modelling the state of every component in great detail. The physical
variables of the entire trajectory from the environmental impact to the stress range
cycles are followed for numerous subsequent time steps until sufficient statistical
convergence is achieved.
118
Chapter 6
Frequency Domain
spectral parameters
of environmental state
Time Domain
power spectra
(and coherence function)
e.g. IFFT
random time series
transfer functions
time step integration
IFFT
stress response spectra
stress time series
spectral
estimation
signal analysis
spectral integration
statistical
relations
spectral moments
analytical or empirical
probability density function
(pdf)
characteristic
parameters
( std, Tz, Tp,
bandwidth )
"Fatigue" Domain
Rainflow counting
(RFC)
stress range distribution
Palmgren-Miner damage accumulation,
SN curve
damage, equivalent load
Frequency Domain Approach
Time Domain Approach
Figure 6.5: Fatigue analysis by either time domain approach (thick arrows) or
frequency domain approach (double arrows)
(Dashed lines indicate alternative paths between time and frequency domain.
pdf: probability density function, RFC: Rainflow Counting)
The approach is very popular in wind energy engineering because it is widely proven by
measurements and capable of treating non-linear and unsteady systems. Even nonGaussian input can be considered, which is however not normally done. Fatigue
analyses in the time domain are observed in offshore technology only for systems with
strongly non-linear behaviour, e.g. floating platforms and riser systems.
119
We are using a 4th order Runge-Kutta explicit integration schemes with constant time
step. As a consequence the highest eigenfrequency dictates the length of the integration
time step leading often to a much finer temporal discretisation than required with respect
to the frequency content of the applied input signals.3
Linear spectral analysis
In contrast with the above, the frequency domain approach deals directly with statistics
of loading and response. Transfer functions can be established in different ways by
analytical or numerical linearisation, by evaluation of the response of a ‘black-box’ model
to test signals or by processing the response on normalised excitations with different
frequencies. The statistics of the stress spectra are transformed by algebraic relations
into a probability density function of the stress ranges. The entire process is between 10
and 100 times faster than Monte-Carlo simulations.
The frequency domain approach is the general method for assessment of the fatigue of
offshore structures and much experience exists with its application [6.21].
Application in the field of wind turbine engineering is not new [6.22] but encounters a
number of problems including time variance of the equations of motion, non-linearities,
unsteady phenomena and periodic effects. Furthermore the bandwidth of the
aerodynamic response is so wide that most frequency domain methods generate far too
conservative results for the fatigue damage. Only the Dirlik method is suited in this
respect, which is further discussed in Section 7.2.2.
There is another practical argument against the frequency domain approach. The
price/performance ratio of current computer hardware is so low that fast and accurate
design calculations in the time domain are viable for onshore wind energy converters.
Preferred approach
For accurate analysis of the entire OWEC or the wind turbine part the time domain
approach is currently regarded as the only suitable choice. However, the frequency
domain approach is certainly appropriate for the support structure under pure
hydrodynamic loading.
In Chapters 7 and 8, simplified analysis methods for the early design stage are
developed that combine the advantages of both approaches by the combination of time
domain analysis of the aerodynamic response and frequency domain analysis of the
hydrodynamic response of the support structure.
6.3. Model reduction
Reduction of degrees of freedom
6.3.1
Most time domain wind turbine design tools apply a relatively low number of degrees of
freedom only, typically around 20. Only with this number the complexity of the non-linear
equations of motion and the computational time remains reasonably limited.
Computation on typical PC-based platforms results in a duration between several hours
to approximately one day for one entire design calculation of an onshore wind energy
3
For numerical stability the integration time step of the Runge-Kutta scheme must be smaller than 2.8
divided by the maximum eigenfrequency of the system. For accurate results a much smaller time step
is required.
120
Chapter 6
converter comprising between 12 and 25 fatigue load cases and 20 to 200 extreme load
cases.4 Offshore design calculations involve more complex structural and loading
models and a larger number of fatigue load cases. Commonly long random realisations
per load case, of for instance 3 hours, are applied in offshore engineering if time domain
analyses are carried out. In contrast, in the wind energy community only a few number
of simulations of 10 minutes each are used in practice.5 Therefore it is evident that the
computational effort for the analysis of offshore wind energy converters has to be
limited.
In Section 6.1.2 the need for a Finite Element model of the support structure was
explained. The resulting number of physical degrees of freedom, typically between 20
and 600 if beam elements are used, is too high. However, various methods exist for the
reduction of degrees of freedom of structures with stiffness or mass proportional
damping, linear kinematics and linear constitutive relations. The most important for our
purposes are the modal reduction or mode superposition and the static or Guyan
reduction.
Prior to the application of any of the two methods, the complexity of the model should
be chosen appropriate to the support structure type, the features of the simulation code
and the analysis type. Relatively few beam elements generally provide a quite
reasonable model and node spacing should reflect changes in the structural properties,
distribution of external forces and hot spot locations. Heave degrees of freedom can
generally be locked. They are not critical but result in unduly high eigenfrequencies.
Consideration of the torsional behaviour is meaningful only if the aero-elastics and the
structural dynamics of the yaw dynamics are modelled in a sophisticated manner.
Mode superposition
The mode superposition is based on the modal decomposition of the system which
transforms the set of n coupled differential equations of the n physical degrees of
freedom, x, into a set of n uncoupled equations for the same number of generalized or
modal coordinates, q. The transposed modal matrix, Φ T, with the eigenmodes, nj j =
1 .. m, constitutes the transformation from the physical to the modal coordinates, (6.1).
Consideration of only m modes, with m smaller than n, generates a reduced set of
generalized coordinates
 q1 
 ϕ 1,1 L ϕ n,1 
 

O
M 
 M  =  M
q 
ϕ 1,m L ϕ n ,m 
 m
 x1 
 
M
x 
 n
m < n
(6.1)
4
Different combinations of mean wind speed, mean wind direction, initial rotor position and system
properties generate a high number of extreme load cases each with relatively short duration.
5
Taking the Opti-OWECS monopile configuration and one severe sea state as example, Hofland [6.23]
investigated the influence of simulation length and number of averaged simulations on different response
properties. Quite similar convergence of the damage equivalent loads was found between separate
simulations of the ware and wind response. For waves the 95% confidence intervals accounted for ±10%
and ±4% when one time series of 10 min and 3 time series of 20 min each were used, respectively.
Slightly lower values of ±7% and ±3% were observed for the wind case. Further research is required to
establish more general conclusions on the required simulation technique.
121
and modal equations of motion with diagonalized matrices for the generalised mass,
mgen, damping, dgen, and stiffness, kgen and the generalized forces, ΦT·f, (6.2).
O
  M  O
  M  O
  M   ϕ11, L ϕ n,1 
 m
 q&&  +  d
 q&  +  k
 q  =  M O M 
gen i
gen i
gen i

  i 
  i 
  i 








O  M  
O  M  ϕ1,m L ϕ n,m 
O  M  
i = 1..m, j = 1..n
M
 
f j 
M
 
(6.2)
This treatment is possible if proportional damping is assumed which is a reasonable
assumption for the material damping of the structure. In most cases one decomposes
the conservative system (without damping) and considers the structural damping as
modal damping by prescribing the generalized damping.
The models for soil damping of piled or gravity foundation, described in Sections 5.2.2
and 5.2.3, result in non-proportional damping that is applied at the mudline. So, the
physical damping matrix is empty except of the few diagonal elements associated with
the concerned degrees of freedom at mudline. Modal decomposition of such a system
results in complex eigenmodes [6.24]. We are interested in the low-frequency dynamics
only on which the soil damping has a rather small influence (Section 9.4.3). Therefore
we avoid such a complication and use the real eigenmodes of the conservative system
for the modal decomposition. As a consequence the few elements of the physical
damping matrix are spread over the entire generalized damping matrix which is still real
but no longer diagonalized. Nonetheless, still a simple and efficient numerical
implementation is possible.
In a certain variant, the so-called ‘mode superposition with static improvement’, the
dynamics of only a low number of modal degrees of freedom, qdyn, with eigenfrequencies
in the range of the excitation are considered, (6.3). The inertia and damping forces from
the remaining higher modes are neglected but the static deflection of these degrees of
freedom, qstat, improves the overall solution.
 m gen dyn




O
 q&&dyn   d gen dyn

+
 O  


O
 q& dyn   k gen dyn

+
 O  


k gen 
stat 
T 
 q dyn   Fdyn

 =  T  {f } (6.3)
 q stat   Fstat 
Three criteria govern the selection of modes. Firstly, the range of the associated
eigenfrequencies should cover the frequency content of the input signal. Too high
eigenfrequencies would require an uneconomically small integration time step because
of the stability requirement of the explicit integration scheme with a constant time step.
Secondly, only such modes can be excited effectively whose shapes show significant
deflection at the application point and in the direction of the loading. Finally, if the static
response and the internal member forces are important, a relatively high number of
modes might be worthwhile.
Sometimes, difficulties occur owing to the mode superposition method during the
calculation of member forces in complex structures, e.g. of the lattice type. The required
superposition of the 3rd and 4th order derivatives of the displacement field converges only
poorly or not at all, even for a very high number of modes. Therefore the common choice
in the offshore industry for the analysis of jacket platforms is the direct solution in the
physical coordinates rather than mode superposition [6.25].
122
Chapter 6
Static reduction
For the static reduction after Guyan [6.26] the equations of motion are partitioned
according to a reduced number of master degrees of freedom, xM, and the remaining
slave degrees of freedom, xS , (6.4).
 M MM M MS   x&&M   DMM DMS   x& M   K MM K MS   x M   fM 
(6.4)
M
  +
  +
  = 
 SM M SS   x&&S   DSM DSS   x& S   K SM K SS   x S   fS 
The identity relation and the static equilibrium between master and slaves establish the
transformation matrix, T, (6.5).
 xM 
 = T
 xS 
{x M }
where:
I


T = 

−1
 − K SS K SM 
(6.5)
Multiplication with the transposed transformation matrix from the left and the
transformation matrix from the right obtained the reduced equations of motion which are
not uncoupled but have considerably smaller the size.
M red x&&M + D red x& M + K red x M = T T f
For illustration we write out the reduced mass matrix, (6.7).
[
-1
-1
-1
M red = M MM - M MS K SS
K SM - K MS K SS
M SM - M SS K SS
K SM
(6.6)
]
(6.7)
Opposite to the modal reduction, this approach is exact for static loading but
approximative for the dynamic case. Master degrees of freedom should be selected in
such a way that the shapes of important eigenmodes can be reproduced reasonably and
that master degrees of freedom correspond to the location and direction of the loading.
General purpose Finite Element codes provide also an opportunity for the automatic
master selection based upon the associated mass - stiffness ratio. Good results are
achieved by a combined approach of manual selection of the majority of master degrees
of freedom by the analyst and automatic selection of the remaining portion of, say, 10 20%.6
Implementation in the DUWECS code
The support structure module within the DUWECS simulation tool always applies mode
superposition. However, several variants are possible according to the particular needs:
• mode superposition without modal reduction
This unusual choice uses all eigenmodes that are extracted by the preceding Finite
Element analysis. It is suited only for very simple models without high
eigenfrequencies or after a static reduction.
• modal reduction
Eigenmodes are chosen according to the three criteria mentioned above. Different
numbers of modes should be tested in order to check the desired accuracy.
6
For the present case of coupling a linear elastic model of the support structure to a non-linear model
of the rotor, a mixed static and modal reduction of the degrees of freedom [6.27] would be an ideal
approach. It was not applied here due to the cumbersome implementation.
123
• mode superposition with static improvement
This recommended option integrates several advantages. The dynamics are well
represented by a low number of dynamic degrees of freedom. A coarse time
integration step is applicable because high eigenfrequencies are not relevant. The
static solution and especially the member forces are improved by the contribution of
a larger number of static modal degrees of freedom for which the inertia and damping
forces are neglected.
• static reduction prior to any of the mode superposition variants
The support structure model itself is derived with aid of the general purpose Finite
Element code ANSYS 5.3™ which enables a static reduction of the model prior to the
DUWECS input. Thus, a static reduction can be combined with any of the abovementioned methods.
Calculation of member forces by mode superposition
max. rel. error
w.r.t. direct solution in physical coord.
The suitability of the mode superposition technique was investigated for two typical
support structures limiting the entire range of possible solutions, i.e. a tubular
monotower and a fully braced lattice tower (Figure 9.4).
In the case of the monotower, classical mode superposition was found sufficiently
accurate by selection of the ten lowest bending modes per direction (Figure 6.6). The
associated eigenfrequencies up to 70 Hz require, however, an uneconomically small
time integration step. So one has to accept a lower accuracy by using only the first five
modes up to 10 Hz if classical mode superposition is desired. As above-mentioned,
mode superposition with static improvement is of great advantage since only the first
three to five modes are needed for the descriptions of the dynamic response whilst the
10.00%
Gravity monotower:
excitation due to harmonic thrust
.
force at 0.9 f 0 , 1.5% damping
1.00%
0.10%
pure mode superposition
with static improvment (10 total modes)
with static improvement (20 total modes)
0.01%
0
2
Mode 1:
0.52 Hz
4
6
Mode 5:
10.6 Hz
8
10
12
Mode 10:
71 Hz
14
16
18
20
22
Mode 20:
246 Hz
number of considered fore-aft modes as dynamic degrees of freedom [-]
Figure 6.6: Convergence of member forces at a tubular support structure calculated
by mode superposition compared to mode superposition
with static improvement
124
Chapter 6
rest are static degrees of freedom without negative consequences for the time
integration step.
Proper modelling of lattice type structures required more effort. A static reduction with
about 50 - 100 master degrees of freedom is recommended. Subsequent mode
superposition with static improvement showed similar results as for the monotower;
however, a considerably higher number of modes was required to achieve a reasonable
accuracy for the member forces.
significant wave height [m]
6.3.2
Lumping of load cases for wind and wave fatigue
For an offshore wind energy converter at least an order of magnitude more fatigue load
cases exist than for an onshore wind energy converter. Consideration of directional
effects, typically done by 12 sections, increases the number of load cases by another
order of magnitude. Evidently this is incompatible with integrated time domain
simulations and establishment of a low number of characteristic load cases is required.
A so-called ‘lumping of load cases’ is well known in the offshore technology for pure
hydrodynamic loading.7 We have to deal with significant wind and wave fatigue as well
as the dependency of the response behaviour on the aerodynamic damping for different
wind speeds and operational conditions.
In Chapter 8 a simplified analysis approach for support structure fatigue owing to wind
and wave loading is introduced. This approach enables a very fast calculation of
hundreds of load cases which provides the basis for the lumping of sea states. The
environmental parameters of a lumped load case are selected in such a manner that
they will result in approximately the same damage that accumulates considering all
elementary load cases with spread environmental parameters.
'damage
centre'
6
area represented by
a lumped load case
5
4
3
2
1
0
0
5
10
15
20
25
30
hourly wind speed at 60 m height [m/s]
Figure 6.7: Scatter of 58,333 sea states at the NL-1 site during 9 years.
(Boxes include such sea states that are considered in the 21 lumping load cases from
Table 8.3. The damage centres indicate the parameters of the lumped load cases.)
7
In structural dynamics, `lumping` has a quite different meaning and denotes the fictitious concentration
of a property, for instance the structural mass, in the nodes as an approximation of the continuous
distribution in reality.
125
Each dot in Figure 6.7 represents one sea states occurring during nine years according
to the NESS database and was grouped into a three-dimensional scatter diagram with
a minimum probability of 0.5 0/000. From the 235 elementary load cases of the scatter
diagram 21 lumping load cases were constructed which parameters or ‘damage centres’
are indicated by the markers.
The details of the lumping approach are explained in Section 8.3 where the approach
is developed in the context of the other simplified methods.
Pre-Processing
Design Data
Offshore Wind Energy Converter
Environmental Parameters
ANSYS
(Finite Element Code)
Support Structure
and Soil Model
SWING
(Stochastic Wind
Field Generator)
Rotor, Transmission,
Generator and
Controler Model
Simulation
DUWECS
(Integrated, Non-Linear, Time-Domain Simulation
of Aero-Elastic, Structural and Hydrodynamic Behaviour)
Loads, Displacements,
Member Forces
Post-Processing
FAROB
(Strength and Fatigue Analysis)
Extreme Stresses, Fatigue Damage
Figure 6.8: Implementation of simulation approach in the time domain
6.4. Implementation
In this thesis a particular implementation of the integrated, modular, non-linear time
domain approach has been developed and implemented by major extension of the basis
design tool DUWECS (Delft University Wind Energy Conversion Simulation) which
originally included only a single degree of freedom model of onshore towers [6.28].
Details of the modelling are discussed in Chapter 5.
Figure 6.8 explains the general analysis process. The total simulation process is divided
into three parts with respect to the chronological order and the software used. The basic
126
Chapter 6
idea of this approach is to minimize the computational effort in the actual simulation by
treating as many as possible tasks prior or after the main process.
• Pre-Processing (model generation and reduction, wind generation)
The general purpose Finite Element code ANSYS 5.3™ is used for the generation
and reduction of the model of the support structure including the non-linear
foundation. A modal analysis with or without pre-stress effects establishes the model.
Geometric data, the eigenmodes and eigenfrequencies are transferred in an object
orientated format [6.29] by an interface program [6.30] which also generates modal
unit load vectors and modal unit member forces.
The spatial and temporal structure of the wind field is calculated by the SWING code
(Stochastic WINd Generator) [6.31]. Furthermore, input data for other modules have
to be prepared by the user.
Various utilities are available for generation of input files and processing of batch
calculations.
• Simulation (response calculation)
Time histories of displacements, loads and member forces are computed with the
simulation code. Aerodynamics is based upon standard blade element momentum
theory. Wave kinematics of regular or random waves according to non-linear Stream
function wave theory or linear wave theory with Wheeler correction are determined.
The hydrodynamic forces on structural members with arbitrary orientation are
calculated with the Morison equation optionally corrected for wave diffraction.
In addition, DUWECS is capable of linearizing the system at the equilibrium state.
The calculated state-space model can be used for frequency domain analysis,
controller design, calculation of the coupled eigenfrequencies and determination of
the aerodynamic or soil damping. Hydrodynamic transfer functions can be generated
by steady state simulations of the response to elementary waves.
• Post-Processing (response analysis and interpretation)
Signal analysis and graphics are performed by another software package, e.g.
MATLAB™. The fatigue analysis with Rainflow-Counting and linear Palmgren-Miner
rule is done in the FAROB code which has been developed at the Stevin Laboratory
of the Delft University of Technology [6.32]. Again utilities are provided for batch
processing.
Bases for Simplified Fatigue Analysis Approaches
127
CHAPTER 7
BASES FOR SIMPLIFIED FATIGUE ANALYSIS APPROACHES
Application of the integrated, modular, non-linear time domain simulation approach,
described in the previous chapter, can potentially lead to more cost-effective and
reliable design solutions for offshore wind energy converters. The high computational
effort associated with the fatigue analysis of the simultaneous aerodynamic and
hydrodynamic response of the support structure in the time domain is, however, not
compatible with the iterative nature of the early stages in the design process. In such
situations it would be convenient to separate wind and wave response since wellestablished and fast approaches exist in both wind energy and offshore technology to
deal with either wind or wave response.
Indeed it is possible to establish such simplified (but not simplistic) approaches. This
chapter demonstrates how the results of separate analyses of wind and wave induced
fatigue can be combined in a convenient way which accounts for the two important
effects of partial cancellation of wind and wave responses and aerodynamic damping
of the wave response. Figure 7.1 compares a simplistic superposition neglecting both
aspects to one of the new simplified approaches for the equivalent bending moment at
the mudline of the Opti-OWECS monopile.
Before the two novel methods are demonstrated in a cookbook manner and on an
200%
damage equivalent load
175%
aerodynamic damping
150%
125%
cancellation
75%
waves without
aerodynamic
damping
50%
wind
100%
25%
0%
simplistic
superposition
simplified OWEC
analysis
combined
response
Figure 7.1: Simplistic superposition of separate wind and wave fatigue responses
versus simplified analysis approach
128
Chapter 7
empirical basis for the long-term fatigue loads in Chapter 8, we have to establish a more
theoretical basis here. Therefore, this chapter deals only with the short-term statistics
of stationary processes.
The problem is split up into two major aspects: Separation of the simultaneous response
in Section 7.1 and superposition of separated responses in Section 7.2.
Evaluation of the interactions within an offshore wind energy converter (OWEC) in
Section 7.1.1 reveals the aerodynamic damping of the fore-aft motion as essential for
our purpose. This insight enables the separation of the simultaneous response into an
aerodynamic and hydrodynamic contribution (Section 7.1.2). Different estimates of the
aerodynamic damping are evaluated in Section 7.1.3 while the substitution by additional
structural damping is demonstrated in Section 7.1.4.
Next, the superposition of fatigue loads is elaborated. After comparing different
superposition strategies in Section 7.2.1, the chosen combination of damage equivalent
fatigue loads is derived and validated in Sections 7.2.2 and 7.2.3, respectively.
The reference case for the numerical investigations in this chapter is the design solution
of a 3 MW, two-bladed, pitch regulated offshore wind energy converter with constant
rotor speed and a soft-soft monopile developed during the Opti-OWECS study (Figure
9.4, Section 9.1). In contrast to Chapter 9, here only a simplified model of the support
structure (Figure 7.2) is taken from reference [7.1]. The lowest five fore-aft bending
modes are used. Wave loads rest on the linear wave theory in its original form.
Hydrodynamic drag is not significant under fatigue conditions and is neglected in the
frequency domain analyses. The wind turbine model is identical to that from Chapter 9.
Further data can be found in Appendix B.
In this chapter we restrict us to collinear and omnidirectional waves which results for the
reference design in conservative fatigue loads compared to situations with a
misalignment between wind and wave direction (Section 9.6.1) and a directionality of
wind and wave loading. So, the influence of the wave excitation on the lateral response
is not significant, pure aerodynamic response dominates in the side direction and the
computational effort is limited by considering only the fore-aft response.
Morever we restrict the following derivation to fatigue caused by production load cases.
Parked and transitional situations, where either wave or wind loads dominate, are
discussed in Chapters 8 and 9.
7.1 Separation of simultaneous response under wind and wave loading
7.1.1 Independence of wind and waves and dynamic interaction of responses
The underlying idea for the separation of the OWEC response is that wind and wind
generated wave loading can be regarded as independent and stationary on a short time
scale, i.e. within a period of ten minutes to three hours. Figure 7.3 plots two time series
as an example. The process of wave generation by the wind is driven only by the mean
wind speed rather than by atmospheric turbulence.1
1
The wave phase velocity (celerity) is slower than the mean wind speed, so wave crests and troughs are
not synchronised with atmospheric eddy structures. Local influences of the instantaneous water surface
elevation on the wind field are relevant only in the splash zone but not at the swept rotor area, located at
least 10 - 30 m above the sea level.
129
Non-linear constitutive relations and geometric non-linearities are not significant for
fatigue conditions. Consequently time series and spectra of the quasi-static responses
of independent, random aerodynamic and hydrodynamic loading processes can be
superimposed from separate analyses.
As the aerodynamic forces depend on the structure’s kinematics, superposition of the
dynamic responses must account for the mutual interaction and superposition is not
straightforward.
Figure 7.4 shows the associated bending moments at mudline as an example for the
interaction of the dynamic responses. Superposition of the separate responses
calculated without aerodynamic damping of the fore-aft motion would significantly
overestimate the combined response.
141 t, 8.1•106 kgm2
Ø 2.8 m x 20 mm
40 m
Loc. 7
58.6 m
48 m
Ø 2.8 m x 28.5 mm
14.1 m 4.5
m
Loc. 6
24 m
Ø 2.8 m x 60 mm
MSL
Loc. 5
21 m
0 m (MSL)
Ø 3.5 m x 75 mm
Loc. 4
-12 m
fixity length = 4.6 D
Loc. 3
-21m
Loc. 2
-29 m
Loc. 1
-37 m
Figure 7.2: Opti-OWECS monopile support structure (left) and simplied model used
within this chapter (right)
The various dynamic interactions within the OWEC and between the converter and both
environment and grid are illustrated in Figure 6.1. Not considered are stationary effects
with respect to the duration of the random loading process, for instance, wind generated
130
Chapter 7
waves or the influence of the roughness of the sea surface on the atmospheric
turbulence and wind shear.
For the present discussion the following four aspects are relevant:
•
interactions between the dynamics of the OWEC components,
•
aero-elastics, especially aerodynamic damping,
•
interaction between the hydrodynamic forces and structural response,
•
soil-structure interaction.
Interactions between the dynamics of the OWEC components
The dynamic behaviour of rotor, conversion system including the drive train, support
structure and controller is strongly interrelated with the aerodynamic loading. Since a
couple of years it is common practice to take into account the structural dynamics of the
entire system during the calculation of the aerodynamic response. As an additional
requirement, a time domain simulation approach is preferable with respect to the nonlinear rotor aerodynamics.
Aero-elastic interactions, especially aerodynamic damping
Several aero-elastic effects can be observed at wind turbine rotors and their supporting
structures. Common to all is the feed-back loop between the dynamic behaviour of the
structure, including elastic and inertia forces, the aerodynamic loads and the active
control system (if any). The most important issue for this investigation is the
aerodynamic damping of the fore-aft motion of the support structure motion experienced
during power production of the wind turbine which was already discussed in Section 5.3.
Interaction of the hydrodynamic forces and structural response
Significant interaction between hydrodynamic forces and structural kinematics occurs
only for slender and flexible members, such as cables, moorings and risers. The
alternation of the hydrodynamic drag force owing to the moving structure is experienced
as hydrodynamic damping. For bottom-mounted OWEC support structures the structural
velocities of the submerged part are very small compared to the water particle velocities
and the viscous and radiation damping is much smaller than the structural damping
(Section 9.4.2). So, wave loading, i.e. the external forces, can be calculated separately
from the structural response and the wind loading. From the previous discussion of the
aerodynamic damping, it is obvious that wave response, i.e. the internal member forces
due to wave loading, will be significantly affected by the response of the entire OWEC.
Soil-structure interaction
The magnitude of the stress ranges in the soil due to either aerodynamic or
hydrodynamic fatigue loading is small compared to the ultimate strength. Hence soil
behaviour can be regarded as linear or can be linearized with respect to the stationary
aerodynamic and gravitational pre-loading. Separate analyses of the wind and wave
responses with the same soil model are therefore suitable.
131
Wind speed (rotating observer, spatially averaged, Vmean = 8.1 m/s)
10
[m/s]
9
8
7
6
[m]
Water surface (Hs = 1.1 m, Tz = 4.3 s)
0.5
0.25
0
−0.25
−0.5
120
130
140
150
time [s]
160
170
180
Figure 7.3: Example time histories showing wind and wave loading
to be uncorrelated on short time scale
Separate responses: solid = wind, dashed = waves without aerodyn. damping
15
[MNm]
10
5
0
−5
Combined wind & wave response: solid = with, dashed = without aerodyn. damping
[MNm]
20
15
10
5
0
120
130
140
150
time [s]
160
170
180
Figure 7.4: Example time series of bending moment at mudline for separate
wind and wave responses (upper) and combined responses with and
without aerodynamic damping (lower)
132
Chapter 7
7.1.2 Contributions of wind and waves to the simultaneous response
Separation of the contributions of wind and waves to the simultaneous OWEC response
is far from simple since the aerodynamic loading is affected by the hydrodynamic
response and vice versa.
Here a practical solution is proposed for the development of a simplified analysis
approach. It is shown that the aerodynamic contribution is well approximated by
simulation of the response under wind loading in a calm sea. Substitution of the
aerodynamic damping by an addition to the modal structural damping of the
fundamental fore-aft mode (Section 7.1.4) is a reasonable approach determining the
hydrodynamic contribution.
Approximation of the wind contribution
Even in the case of significant hydrodynamic excitation the magnitude of the tower top
velocity is relatively small compared to the variation of the longitudinal wind velocity.
Likewise the difference of the thrust loading between the simultaneous wind and wave
response and the fictitious case of pure wind loading in a calm sea is small in relation
to the magnitude of the thrust itself. The observation is illustrated by time series of two
load cases in Figures 7.5 and 7.6.
Figure 7.5 compares the response for simultaneous wind and wave loading and pure
wind loading in a calm sea at the partial load conditions from Figure 7.3. The upper plot
shows the thrust loadings at the hub for pure wind loading and simultaneous wind and
wave loading. Both signals are very similar, so, the their difference is given by the time
series in the lower diagram. In addition, the difference of the associated fore-aft velocity
of the tower top is included. Assuming linear behaviour the difference of the thrust
forces can be regarded as the aerodynamic damping force related to the wave
contribution of the simultaneous response. Apparently, this force is negative and
proportional to the difference of the tower top velocities.
The employment of the pitch controller at full load is demonstrated at a higher mean
wind speed of 21 m/s and a corresponding severe sea state (Figure 7.6). Above the
rated wind speed the analog PI-controller regulates the blade pitch angle to minimize
generator speed variation (Figure 5.6, Table B.7). The asynchronous generator with
rated 3% slip is directly connected to the grid which ensures a stiff speed-torque
characteristics. The maximum pitch rate accounts for 5 o/s.
The upper two subplots and present the wind and wave loading and responses of the
pitch controller and the aerodynamic thrust for simultaneous wind and wave excitation.
The two lower subplots and show the difference of the time series associated with the
simultaneous response and the pure aerodynamic response. The relation between
tower top velocity and thrust force variation in the third subplot is similar to the abovementioned partial load situation. The pitch controller is, however, sensitive to the wave
excitation. Waves induce a tower top response which affects the aerodynamic forces
and thus also rotor torque and rotor speed. The deviation of the rotor angular velocity
triggers the controller. Comparison of the difference of the pitch angle variation and the
difference of the tower top velocity in fourth subplot reveals an undesired effect of this
rotor speed controller. Instead of showing synchronous behaviour compared to the wave
induced tower top velocity, the pitch difference has a phase delay of approximately 90o,
which results in a reduced aerodynamic damping.
133
wind [kN]
wind
200
175
200
150
175
wind & waves
150
wind & waves [kN]
Thrust loading at tower top Fx
Velocity difference ∆dxttop , thrust difference ∆Fx: wind & wave − wind response
4
∆F
0.1 ∆dx
x
2
0
0
∆dx
ttop
ttop
−0.1
−2
−0.2
120
130
140
time [s]
∆Fx [kN]
[m/s]
0.2
−4
160
150
Figure 7.5: Responses at partial load without pitch control:
Aerodynamic thrust of pure wind and simultaneous wind and wave response (upper),
Thrust difference opposite to difference of tower top velocity (lower)
[m/s] ∆dx
ttop
[m/s]
θ [deg]
Pitch angle θ and thrust Fx for wind & wave loading
14
300
Fx
θ
12.5
250
11
200
9.5
8
150
Velocity ∆dx
,
thrust
difference
∆F
:
wind
&
wave
−
wind
response
ttop
x
∆dx
ttop
∆F
0.15
3.5
x
0
0
−0.15
−3.5
η [m]
Fx [kN]
3
1.5
0
−1.5
−3
∆dx
ttop
Velocity ∆dx
, pitch angle difference ∆θ: wind & wave − wind response
ttop
∆dx
∆θ
ttop
0.15
0.05
0
0
−0.15
−0.05
120
125
130
135
140
time [s]
145
150
155
x
η V
mean
∆F [kN]
24
21
18
15
12
∆θ [deg]
V [m/s]
Wind and wave loading (Vmean = 21 m/s, Hs = 4.35 m, Tz = 6.5 s)
160
Figure 7.6: Responses with pitch control at full load: Phase shift of differential
pitch response with respect to differential tower top velocity in plot (d)
134
Chapter 7
For 17 production load cases similar to Table 8.3 the thrust loading is compared
between pure wind response and simultaneous wind and wave response. The
difference of the aerodynamic forces was found large enough to employ effectively the
aerodynamic damping influence on the wave induced response, but it was too small to
significantly affect the wind induced fatigue loads. The aerodynamic thrust of the pure
aerodynamic response underestimated the thrust associated with the simultaneous wind
and wave response by 1.5% ±0.4% in terms of damage equivalent fatigue loads with
respect to an S-N curve with inverse slope of four.
In summary, the wind contribution to the combined response can be approximated by
the wind response in a calm sea, taking water added mass as the only hydrodynamic
effect. This corresponds to the application of an onshore wind turbine design tool where
extra mass at the tower base is considered.
Approximation of the wave contribution
The establishment of the wave contribution to the combined response is more involved
due to the aerodynamic damping. Three possible approaches are identified.
One may define the wave contribution as the remaining share after subtracting the wind
contribution from the simultaneous response [7.1]. This leads to difficulties because the
wind contribution has only been approximated and the combined response is available
only from a time-consuming analysis with the integrated OWEC model.
Secondly, the wave contribution is simulated by the wave response of the OWEC
operating in a steady wind field. Despite the good approximation (Section 7.1.4) and
previous work of the author this approach is not further recommended owing to the
requirement of an integrated dynamic model of the OWEC.
Following an engineering approach, aerodynamic damping is substituted by increasing
the modal damping ratio of the structure’s fore-aft mode. No integrated model is needed
and any design tool for offshore platforms should be suitable to calculate the wave
response. The difficulty lies, however, in the consideration of the parameters and
operational conditions that influence the aerodynamic damping. Nonetheless this
approach is preferred and investigated in the next two sub-sections.
7.1.3 Estimation of the aerodynamic damping
Determination of the aerodynamic damping for certain values of the mean wind speed
(or non-dimensional tip speed ratio) and pitch angle is far from new. Kaiser, for instance,
provides a comprehensive overview [7.2]. In the design practice such investigations are
rare since common time domain design tools are not capable of a linearisation of the
wind turbine model. Therefore we analyse the suitability of three methods for the
approximation of the aerodynamic damping of the fore-aft translation in a separate
analysis of the wave response.
Closed-form linearisation
The simple, closed-form model of the linearized aero-elastic behaviour from Eqn. (5.12)
enables physical insight and the identification of important parameters. According to
Prandtl the slope of the lift coefficient equals approximately 2@π for attached flow and an
upper bound of the damping can be established easily.
Information on the stationary flow conditions along the rotor blades is required for more
realistic conditions with partially separated flow. Any standard wind turbine code or even
135
Table 7.1: Example of stationary flow conditions used as input
for the closed-form estimate of aerodynamic damping (5.12)
Initial inflow conditions with uniform wind field
i.e. without blade velocity, wind shear, yaw, tower shadow, etc.
Partial load
Undist. veloc.
[m/s]=
13.0000
Induction factor
[1]=
.2088
(1-a)*Vw0
[m/s]=
10.2859
Rot. speed
[rad/s]=
2.3038
Theta,root,rated [deg]=
13.4000
Theta,controller [deg]=
.0000
Theta,root
[deg]=
13.4000
air dens. roh [kg/m^3]=
1.2500
kin.visc. eta [kg/(ms)]= .1820E-04
i
[-]
1
2
3
4
5
6
7
8
9
10
11
r(i)
phi alpha
cl
cl'
cd
cd'
W
Re
[m] [deg] [deg]
[-]
[-]
[-]
[-] [m/s]
[-]
6.0 36.7 23.3 .122E+01 -.103E+01 .342E+00 .159E+01 17.2 4.5E+06
9.6 25.0 14.2 .147E+01 -.195E+01 .945E-01 .160E+01 24.4 7.2E+06
13.0 19.0 11.2 .153E+01 -.344E+00 .316E-01 .573E+00 31.7 9.4E+06
16.4 15.2
9.8 .146E+01 .301E+01 .186E-01 .527E+00 39.1 9.9E+06
19.7 12.7
8.8 .140E+01 .564E+01 .113E-01 .103E+00 46.6 9.5E+06
23.1 10.9
8.0 .133E+01 .564E+01 .988E-02 .103E+00 54.3 9.2E+06
26.5
9.6
7.6 .130E+01 .576E+01 .103E-01 .859E-01 61.9 8.9E+06
29.9
8.5
7.4 .128E+01 .576E+01 .101E-01 .859E-01 69.6 8.3E+06
33.3
7.6
7.5 .131E+01 .588E+01 .115E-01 .659E-01 77.3 7.6E+06
36.7
6.9
7.5 .131E+01 .588E+01 .116E-01 .659E-01 85.1 6.7E+06
40.1
6.4
7.4 .129E+01 .588E+01 .114E-01 .659E-01 92.8 4.5E+06
a spreadsheet based model of the elementary blade element momentum theory can be
applied for this purpose. For illustration, we evaluate a condition just below the rated
wind speed with the DUWECS simulation code (Table 7.1). The angle of attack (fourth
column) at the inner four blade sections is large enough that partial flow separation and
reduction of the slope of the lift coefficient (sixth column) occurs. Eqn. (5.12) estimates
2.9% of the critical damping. In comparison 5% are obtained for fully attached flow at
wind speeds below 9 m/s.
Numerical linearisation
A numerical linearisation is capable of more comprehensive rotor aerodynamics,
unsteady effects, a larger number of degrees of freedom and the dynamics of the
control system. In general such an option is not implemented in standard wind turbine
design tools used for dynamic load calculations. So, one has to use specific aero-elastic
codes [7.2] or extended design tools.
In Section 5.1.5 it was explained how the simulation code DUWECS is able to derive a
linear state space model of the entire OWEC (5.1). The aerodynamic damping can be
calculated conveniently from the complex eigenvalues of the state matrix of the
uncontrolled system, A (5.2), or from the state matrix, Aclosed-loop (5.4), if the controller
dynamics are included.
Non-linear time domain simulation
A more cumbersome, but easier available approach applies non-linear time domain
simulations. Different methods are available, e.g. evaluation of the steady state
response under harmonic excitation, analysis of the transient decay of free vibrations
136
Chapter 7
after a step or pulse loading or response spectra analysis. In principle such treatments
should be possible with any standard wind turbine design tool and some manual postprocessing.
Here transient vibrations were analysed to study the influence of the vibration amplitude
and inflow. Two alternatives are discussed in more detail for a mean wind speed of 12
m/s. In the first case a thrust pre-loading of 10% of the rated loading was released after
30 seconds and free vibrations in steady inflow were observed (Figure 7.7).
The free vibrations were also excited in the presence of a turbulent wind field (Figure
7.8). The deterministic vibrations and the stochastic response are of the same order of
magnitude. The contribution of the free vibrations is clear but the damping cannot be
derived unless the stochastic part of the response is filtered out. Assuming the validity
of linear superposition of both contributions, the tower top displacement from a
stochastic simulation with the same random seed but without pre-loading is subtracted
from the combined response. The remaining signal in the fourth plot is regarded as the
deterministic part of the free vibrations in a turbulent wind field.
The logarithmic decrement of the total structural and aerodynamic damping, Λ, is easily
read from the semi-logarithmic plot of the sequence of the n successive maxima a1 to
an (Figure 7.9).
Λ
ξ
'
a
1
ln 1
an
n&1
Λ
'
4 π2 % Λ 2
ξ«1
.
Λ
2 π
(7.1)
The damping ratio, ξ, was not affected significantly by the vibration amplitude. Both
below and above the rated wind speed slightly lower damping was observed for small
vibration amplitudes in a turbulent wind field.
Comparison of different approaches
The three methods, closed-form expression, numerical linearisation and non-linear
simulations, were compared for the reference design. In the transient approach six
turbulence representations for each of the 20 wind speed classes were averaged. Figure
7.10 shows estimates of the aerodynamic damping of the first fore-aft mode against
mean wind speed. Only the stationary but not the dynamic behaviour of the controller
is dealt with. The aerodynamic damping remains constant at a high level up to 10 m/s
since the flow is fully attached. Damping is decreasing until the rated wind speed of 13.7
m/s caused by partial flow separation at the inner blade sections. Beyond the rated wind
speed the pitch controller decreases the angle of attack to limit the power. Flow
becomes more attached and the aerodynamic damping increases again. All three
methods show quite similar results. The slightly lower damping estimates ofe the closedform expression (5.12) are probably caused by neglecting the contribution of the drag
force.
Figures 7.11 and 7.12 highlight the importance of the dynamics of the control system.
In the range just above the rated wind speed the aerodynamic damping increases
compared to the case with stationary controller behaviour. However, for higher wind
speeds the damping lowers dramatically down to only 0.5% at the cut-out wind speed.
The time series in Figure 7.6 illustrate this effect in the time domain.
Wind loading V and thrust pre−loading Fx
40
13
30
12
20
11
10
10
0
Aerodynamic thrust loading Fx
375
350
325
x
F [kN]
Fx [kN]
14
V [m/s]
137
300
275
Tower top displacement x
ttop
[m]
0.55
x
ttop
0.5
0.45
0.4
25
30
35
40
45
time [s]
50
55
60
Wind loading V and thrust pre−loading Fx
14
13
12
11
10
40
30
20
10
0
Fx [kN]
V [m/s]
Figure 7.7: Determination of aerodynamic damping from
transient vibration in a steady wind field
Fx [kN]
Aerodynamic thrust loading F
x
375
350
325
300
275
Tower top displacement xttop: solid = pre−loaded, dashed = not pre−loaded
xttop [m]
0.5
0.4
0.3
∆xttop [m]
Difference of tower top displacements ∆xttop
0.05
0
−0.05
25
30
35
40
45
time [s]
50
55
60
Figure 7.8: Determination of aerodynamic damping from transient vibration in a
turbulent wind field
138
Chapter 7
log (modal vibration
amplitude)
100
10
seed 1
seed 2
seed 3
steady wind
1
30
35
40
seed 4
seed 5
seed 6
45
50
55
60
65
70
75
time [s]
Figure 7.9: Determination of the logarithmic damping decrement of the
first mode from free vibrations at a mean wind speed of 12 m/s
The controller is switched on and off at the rated wind speed. Transition phenomena
cause lower damping in the more realistic turbulent case than for both numerical
linearisation and transient analysis with steady inflow. Above 16 m/s the damping
derived in turbulence is approximately 0.5% higher than in a steady wind field.
The actual behaviour may be specific to the reference case. For instance, introduction
of a threshold for activating the pitch variation might reduce the undesired feed-back
from the wave excitation to the pitch system (Figure 7.6). Nonetheless, the inclusion of
the controller dynamics is essential for the damping of controlled turbines.
The simple PI controller was designed as single input - single output system to limit the
variation of the rotor speed and corresponding generator power. In fact in some
advanced wind turbines the control system fulfills multiple control goals, e.g. reduction
of both power variations and fatigue loads [7.3]. Application to OWEC should be
possible in principle, as well.
Recommendations
A number of conclusions are drawn from the investigations. It should be kept in mind
that only one particular design was analysed. Pitch regulated turbines with variable
speed, for instance, might show different behaviour. The application of either the
numerical linearisation or the transient vibration in a steady wind field is recommended
for the determination of the additional structural damping. Evaluation of transients
should be possible with any wind turbine design tool.
The accuracy of the estimates by the closed-form expression depends strongly on those
of the assumed flow conditions along the rotor blades. If input data from a standard
aerodynamic code are applied, the method might be suited well for the wind speed
range without controller activity. Stall regulated machines should be treated with care
since the effects of aerodynamic drag and unsteady aerodynamics are not included in
the formula.
The numerical linearisation and the free vibrations analysis in steady inflow are
preferred with regard to the more cost and cumbersome free vibrations in turbulence.
The differences between the approaches are not significant with respect to other
inaccuracies of the entire simplified fatigue analysis approach. Generally the accuracy
of any theoretical estimate of the aerodynamic damping is at least in the order of ±0.5%
139
6%
closed-form (without
controller dynamics)
damping ratio [-]
5%
4%
numerical
linearisation
(without controller
dynamics)
3%
2%
1%
0%
5
10
15
20
25
free vibrations in
steady wind (without
mean wind speed at hub [m/s]
Figure 7.10: Aerodynamic damping derived without controller dynamics
damping ratio [-]
6%
numerical
linearisation
(without controller
dynamics)
5%
4%
numerical
linearisation (with
3%
2%
1%
free vibrations in
turbulence (with
0%
5
10
15
20
25
Figure 7.11: Aerodynamic damping derived by numerical linearisation or
free vibration analysis
6%
free vibrations in
steady wind (without
damping ratio [-]
5%
4%
free vibrations in
steady wind (with
3%
2%
1%
0%
5
10
15
20
25
free vibrations in
turbulence (with
Figure 7.12: Aerodynamic damping derived from free vibration analysis
with and without controller dynamics
140
Chapter 7
to 1%. It seems more appropriate to apply a simpler approach and add some
conservatism than to take the risk of introducing modelling errors or larger deviations by
imperfect execution of a more involved approach.
In the transition region of the controller it is recommended to consider the lower
damping estimate from two separate estimates with and without controller dynamics.
7.1.4 Empirical validation of separation of responses
An empirical verification of the separation of the response is carried out in the time
domain. Figure 7.13 illustrates the different approaches used.
First the damage equivalent fatigue loads are calculated after Rainflow Counting (RFC)
of time domain simulations of the simultaneous wind and wave response (top).
In the two separate analysis approaches the time series of the wind response in calm
water are superimposed to those from the wave response which accounts one or
another way for the aerodynamic damping.
In the first case the modal damping ratio of the first fore-aft mode is increased by the
estimated aerodynamic damping. Again damage equivalent fatigue loads are computed
(middle). This methods is applied once with full substitution of the aerodynamic damping
estimated by DUWECS linearisation and once only with substitution of 80% of
aerodynamic damping.
Secondly the wave response in presence of steady wind is simulated with an integrated
model of the OWEC including the dynamic behaviour of the controller. The mean wind
speed was adjusted according to the wave height. The case is considered only for
comparison.
For each load case the same seeds are used for the generation of the stochastic wind
and wave loading regardless whether the simultaneous or separate responses are
simulated. In total four representations each with a duration of 40 minutes were
considered for 17 production load cases similar to Table 8.3. Equivalent fatigue loads
were derived by Rainflow Counting of 4000 to 18500 cycles and for an S-N curve with
an inverse slope of four.
Figures 7.14 and 7.15 show the error of the equivalent fatigue loads at two cross
sections derived from superimposed time series relative to the simultaneous response.
The markers and trendlines present the results for the individual load cases whilst the
bar charts at the right-hand side illustrate the extrapolation to the entire lifetime. The
latter is dominated by the load cases associated with high wind speeds. The two
different cross sections correspond to either a wave or a wind dominated loading
regime.
The scatter between different load cases, probably owing to the random representation
and finite sampling frequency, is significant. Compared to this and the spreading in the
fatigue strength determined by experiments there is only a small difference between the
three discussed approaches. Consideration of the full amount of the estimated
aerodynamic damping appears to underestimate the combined response between 1.5%
and 3.5% with respect to the entire lifetime. Much larger errors occur for the individual
load cases. If only 80% of the aerodynamic damping is accounted for, the combined
responses are approximated conservatively by a small overestimate of the long-term
fatigue loads between 0.5% and 2%.
No doubt the accuracy of the long-term fatigue estimates seems reasonable for a
simplified analysis with any of the two damping cases. Other designs with lower
damping than this pitch controlled OWEC with a soft support structure might be
141
Reference approach
Random wind
Simulation of
wind & wave response
Random waves
RFC
Damage equivalent
fatigue loads
Substitution of aerodynamic damping by structural damping
Simulation of
wind response
Random wind
Additional structural
damping
Random waves
RFC
Damage equivalent
fatigue loads
RFC
Damage equivalent
fatigue loads
Simulation of
wave response
Wave response in steady wind
Random wind
Steady wind
Random waves
Simulation of
wind response
Simulation of
wind & wave response
Figure 7.13: Comparison damage equivalent fatigue loads derived
from integrated time domain simulation (top) and superposition of
separate analyses (middle and bottom)
considerably more sensitive to variations of the assumed damping. As a conservative
measure and with respect to other conditions it is recommended to account only for 80%
of the estimated aerodynamic damping.
The consideration of the aerodynamic damping from simulations of the wave response
in a steady wind field obtained a 1.5% to 3% too conservative estimate of the long-term
fatigue loads. The scatter in the individual results was smaller than for the other two
damping cases. However, this does not justify the considerably higher computational
effort associated with the application of an integrated model.
142
Chapter 7
8.0%
Cross section:
below mudline
superposition:
100% additional
structural
damping
error in equiv. fatigue load [-]
6.0%
4.0%
2.0%
superposition:
80% additional
structural
damping
0.0%
-2.0%
-4.0%
-6.0%
-8.0%
5
10
15
20
entire
25 lifetime
superposition:
aerodynamic
damping from
steady wind
mean wind speed [m/s]
Figure 7.14: Error of equivalent fatigue loads from superimposed time series of
separate wind and wave responses
(Cross section 1 below mudline: Relative contribution of wave loading 71 - 82%)
8%
Cross section:
near tower top
error in equiv. fatigue load [-]
6%
superposition:
100% additional
structural
damping
4%
2%
superposition:
80% additional
structural
damping
0%
-2%
-4%
superposition:
aerodynamic
damping from
steady wind
-6%
-8%
5
10
15
20
mean wind speed [m/s]
25
entire
lifetime
Figure 7.15: Error of equivalent fatigue loads from superimposed time series of
separate wind and wave responses
(Cross section 7 near tower top: Relative contribution of wave loading 19 - 48%)
143
7.2 Superposition of fatigue loading
After establishing the wind and wave contributions to the simultaneous response, the
question arises how to combine the results of separate fatigue analyses. In Section
7.2.1 we analyse different techniques for combining the results of two separate fatigue
analyses. The superposition of damage equivalent fatigue loads is identified as most
suitable. Section 7.2.2 derives two methods on either an analytical or semi-empirical
basis which are numerically validated in Section 7.2.3.
7.2.1 Evaluation of different superposition strategies
Simplistic but conservative approach
A simplistic, too conservative method of combining the results of two separate fatigue
analyses is the in-phase superposition of damage equivalent loads due to wind and
waves (Figure 7.1). In the previous section we have learnt how to consider the
aerodynamic damping of the wave response. Now we are looking for an alternative
superposition technique that accounts for the partial cancellation owing to the random
phase relation between wind and wave response.
A solution was sought systematically in two directions: literature survey and evaluation
of possible combinations during different stages of the fatigue analysis process.
Literature survey
Fatigue due to two independent processes has been addressed by researchers in
different fields, e.g. dynamics of moored floating devices, aerospace engineering,
railway dynamics.
Combined wind and wave fatigue is not significant for bottom-mounted offshore
structures in the oil and gas industry. Support structures are dominated by
hydrodynamic loading. While wind induced fatigue is only important for exposed and
flexible elements such as masts, towers and flare booms. Relevant experience for our
investigation is, however, found from the mooring dynamics of floating devices. The
environmentally induced response can be divided into three frequency ranges, i.e.
steady state response due to current forces, mean wind and mean wave drift forces,
low-frequency vessel motion due to wind and waves excitation of the fundamental
mooring modes and wave-frequency vessel motion in a higher frequency range. Both
types of dynamics introduce fatigue [7.4]. The effect can be approximated well by two
independent narrow banded processes with zero-crossing frequencies differing by a
factor between four and 20.
Lie and Fylling [7.5] compared five approaches which were proposed as alternative to
accurate but expensive time domain simulation and successive Rainflow Counting of
the combined response.
The two authors prefer the ‘combination of two narrow banded Gaussian responses
after Jiao and Moan [7.6]. The fatigue damage due to a slowly varying process is added
to the damage due to the narrow banded wave-frequency process. Where the slowly
varying process consists of the low frequency resonance of the mooring tension and the
envelope of the high frequency wave induced tension.
The approach is founded on a firm theoretical basis but unfortunately it leads to fairly
conservative results in our case since two basic assumptions are not valid. The wind
response of a support structure of an offshore wind energy converter is broad banded
144
Chapter 7
while the wave response is reasonablely narrow banded in most cases. Furthermore,
the responses occur in partially overlapping frequency ranges if the fundamental
eigenfrequency is excited by both wind and waves.
The so-called ‘simple summation’ of the damage of the individual responses produces
highly nonconservative results because of the non-linear dependency of the total
damage on the stress magnitude of the two responses.
In the ‘combined spectrum method’ the fatigue damage is derived from the
superimposed spectra. The broad bandwidth of the combined response produces
conservative results if the damage is calculated under the assumption of a narrow
bandwidth. This is also the case for the bandwidth corrections according to Wirsching
and Light [7.7] or Jiao [7.8]. A better alternative, not considered by Lie and Fylling, would
be the treatment of the combined spectra by the semi-empirical approach after Dirlik,
(7.10) [7.9] 2. For us it is however difficult to properly estimate spectra from the signals
of the wind response.
In the field of aerospace engineering Lambert [7.12] proposed an analytical approach
for the fatigue analysis of a single degree of freedom system with combined sinusoidal
and random responses. A typical example includes the gunfire vibration of a jet aircraft.
The method is not considered here because of the restriction on two narrow banded
component responses with the same zero-crossing frequency.
Sakai and Okumara [7.13] developed a semi-empirical combination of the Rayleigh
stress range distributions associated with two narrow banded processes in distinct
frequency ranges. The method was discarded for our purposes because of the broad
banded wind response.
In different technical disciplines sometimes damage equivalent, constant amplitude
fatigue loads with the same number of cycles are combined. If nothing is known on the
phase relation, an in-phase superposition or arithmetic summation of the load
magnitudes provides a conservative estimate for the combined fatigue loading (Figure
7.16 upper). For spatially uncorrelated loads, e.g. the axial and bending loads in a
in-phase or arithmetic superposition
1
1
1
1
1
0
1.5707
3.1414
4.7121
6.2828
+
0
1.5707
3.1414
4.7121
6.2828
=>
0
0
1.5707
1.5707
3.1414
3.1414
4.7121
4.7121
=>
6.2828
6.2828
0
1.5707
3.1414
4.7121
6.2828
-1
-1
-1
-1
-1
90o out-of-phase or quadratic superposition
1
1
1
1
1
0
1.5707
3.1414
4.7121
6.2828
+
0
-1
1.5707
3.1414
4.7121
6.2828
=>
0
0
1.5707
1.5707
3.1414
3.1414
4.7121
4.7121
6.2828
6.2828
=>
0
1.5707
3.1414
4.7121
6.2828
-1
-1
-1
-1
o
Figure 7.16: In-phase versus 90 out-of-phase superposition
of harmonic signals
2
The Dirlik method is summarized for instance in references [7.10] and [7.11] which are easier available
than the original work.
145
member of an offshore platform, the 90o out-of-phase superposition or square root of
the sum of the squares (quadratic summation) of the load magnitudes has been
proposed [7.14] (Figure 7.16 lower). Unfortunately no theoretical or empirical basis for
the latter case was found in the literature.
Superposition during different stages of the fatigue analysis
Considering the fatigue analyses approaches in the time domain and the frequency
domain as parallel processes, as illustrated by Figure 6.5, a combination of results is
possible at four stages:
•
superposition of time series or response power spectra (after transformation to one
domain)
•
superposition of spectral moments (after derivation of ‘pseudo-moments’ from the
standard deviation and counted zero crossings and peaks)
•
combination of stress range distributions
•
combination of damage equivalent stress ranges.
As previously mentioned, any superposition of damages from the separate responses
is senseless. Highly nonconservative results can be the effect, even in case of one
dominant component response.
A systematic investigation of the four strategies for the superposition of wind response
calculated in the time domain and wave response computed in the frequency domain
was carried out in cooperation with Hofland [7.1].
The first option, the superposition of response time series or response power spectra
is a straightforward solution for independent signals. However, inherent inaccuracies are
caused by the spectral estimate of time series of finite length and owing to the
randomness of the time domain representations of the spectra. Considering also the
large amount of data associated with time series, this option is not suited for early
design purposes. Later in this section the superposition of sufficiently long time series
is used as reference solution during the verification of other superposition approaches.
Another possibility is the superposition of the spectral moments and successive
application of a frequency domain relation for the fatigue damage. Under fatigue
conditions it is reasonable to assume the wind and wave responses as Gaussian.
Referring to Rice’s fundamental work on Gaussian processes [7.15] the zeroth, second
and fourth spectral moment, m0, m2 and m4, can be estimated from the standard
deviation, std, the counted zero-crossing period, Tz , and the peak period, Tp , of stress
time series, σ(t), which are sufficiently long from a statistical point of view, (7.2).
m0 ' std ( σ (t ) )
2
std ( σ (t ) )
m2 '
Tz ( σ (t ) )
2
std ( σ (t ) )
m4 '
Tz ( σ (t ) ) Tp ( σ (t ) )
2
(7.2)
The superposition of the spectral moments encounters a practical problem owing to the
generally broad bandwidth of the combined OWEC response and the given set of
spectral moments. A reasonable estimate of the fatigue loading would be possible only
with the Dirlik approach which requires also the first spectral moment. However, an
accurate estimate of the first spectral moment can hardly be gained from finite time
146
Chapter 7
series.3 Application of other, less sophisticated frequency domain methods tends to be
too conservative.
As mentioned above the method after Sakai and Okumara for the superposition of
stress range distribution is discarded since at least the wind response is nonnarrow. Inphase superposition of two stress range distributions is possible after both fatigue load
spectra are portioned in sets with constant stress ranges and corresponding number of
stress cycles [7.17]. However, application as arithmetic superposition would be too
conservative here. After establishing the so-called ‘direct quadratic superposition’ of the
equivalent stress ranges (7.17) in the next section one may propose to combine the
stress range distributions accordingly. This approach is certainly worth of further
investigation, e.g. when dealing with S-N curves with bi-linear slopes, which is beyond
the scope of the present research.
The problem directly leads to the consideration of the superposition of damage
equivalent, constant amplitude fatigue loads with the same reference number of cycles.
Such quantities are a compact and characteristic representation of the fatigue loads
which are obtained conveniently in both time and frequency domain.
For our problem a clear preference we give to the combination of damage equivalent
fatigue loads which is further developed in the next sub-section.
7.2.2 Derivation of the superposition of equivalent fatigue loads
The purpose of this section is to derive a relation for the superposition of the
aerodynamic and hydrodynamic fatigue loads of with two stationary, random processes.
Basic fatigue theory
In this thesis the material resistance to fatigue failure is represented by S-N curves in
combination with the linear damage accumulation after Palmgren and Miner [7.18]. An
alternative, better suited for the assessment of in-service cracks and for the selection
of inspection methodology, is to use fracture mechanics techniques, which are based
on an analysis of the behaviour of the cracks themselves [7.19]. In contrast to the
offshore or aircraft technology such methods are not frequently used in wind energy
applications to date.
(7.3)
N ( ∆ σ ) ' A ∆ σ &µ
For steel, the influence of the mean stress is negligible and the number of endured
cycles, N, of a periodic loading depends only on the stress range, ∆σ, and the two
empirical constants A and µ. In order to derive analytical expressions it is convenient to
assume S-N curves with constant inverse slope µ.
Different S-N curve are used depending on the loading, environment, local geometrical
detail and last not least the applied standard. For instance, Figure 7.17 illustrates three
3
The first spectral moment, defined by (7.11), can be approximated from the integration of the estimated
response spectrum or from the bandwidth parameter after Vanmarcke (7.9), [7.16]. For narrow banded
processes the Vanmarcke bandwidth parameter, δ, is given by the ratio of the standard deviations of the
slopes of the envelope process and the process itself.
Another indirect way is to back iterate the first spectral moment with the aid of the Dirlik approach from
equivalent stress range (7.6) after Rainflow Counting of the signal.
147
log stress range [MPa]
1000.
Type 'A'
µ = 3 and µ = 5
Approximation
of Type 'A': µ = 4
Type 'C': µ = 3 and µ = ∞
100.
Fatigue reference
strength ∆σ R = 100 Mpa
Type 'A'
6
at N R = 2·10 cycles
Type 'B': µ = 3
10.
1.E+04
1.E+05
1.E+062.E+6 5.E+6
1.E+07
log number of endured cycles
1.E+08
1.E+09
log(N) [-]
Figure 7.17: Examples of S-N curves according to Germanischer Lloyd
(The curve with inverse slope of µ = 4 is used as a linear approximation
of the bi-linear curve of Type ‘A’ with µ = 3 and µ = 5.)
curve types proposed by the guidelines of Germanischer Lloyd. Type ‘A’ with bi-linear
inverse slopes of µ = 3 and µ = 5 refers to a mildly corrosive environment or a corrosive
environment where an adequate protection is provided. For a constant stress range
magnitude in a non-corrosive situation a fatigue limit may be assumed by using type ‘C’
curves, whilst for a corrosive environment without protection the inverse slope of µ = 3
is extended according to type ‘B’.
The Type ‘A’ - curve is most appropriate for us and we are using it for analyses on the
certification level (Sections 9.6.1 and 10.3.2) where a numerical evaluation of the S-N
curve is done anyway. In the preliminary design stage and when an analytical evaluation
is desired or when damage equivalent stress ranges are dealt with (Chapters 7 and 8,
Sections 10.1.2 and 10.2.2) the assumption of a constant slope is convenient. In our
case the bi-linear Type ‘A’ - curve is approximated by an S-N curve with the same
reference fatigue strength but constant inverse slope of µ = 4. Figure 7.17 proves this
to be conservative for a wide range of endured number of stress cycles with the
exception
between 2·106 and 2·107 cycles. For stochastic loading where many different stress
range magnitude are occurring this simplification is conservative in most cases.4
4
Taking the stress range distributions from Figures 9.17 and 9.18 as examples the damage is -4% to 14%
higher at the monopile and 51% to 72% higher at the tower if a constant inverse slope of 4 is assumed.
The too low estimation of -4% occurs for loading case No. 2 at the monopile where almost half of the total
damage of D = 5.3 is associated with stress ranges between 50% and 100% of the reference fatigue
strength.
148
Chapter 7
The ratio between the duration of the process, i.e. in most cases the design lifetime,
and the expected time until a failure is denoted as fatigue deterioration or (fractional)
damage, D. Depending on the applied safety philosophy the design damage is set
between 1 and, say, 0.1 corresponding to a time to failure ranging between the service
life and a 10 times longer period.
After Palmgren and Miner the damage equals the summation of the ratio between the
predicted, n(∆σj), and the endured number of stress cycles, N(∆σj), for all stress range
classes, ∆σj.
n ( ∆ σj )
D ' E j
(7.4)
j N ( ∆ σj )
The stress range distribution is gained either from a histogram, n(∆σj), resulting from
Rainflow Counting of the time series or from a probability density function of stress
ranges, p(∆σ), derived from the spectral moments.
Equivalent stress range
Often it is convenient to introduce damage equivalent, constant amplitude loads. For SN curves with a constant slope such an equivalent stress range, ∆σeq, depends only on
the slope parameter, µ, the range distribution of cycles, n(∆σj), and an arbitrary
reference number of cycles. In wind energy engineering, for instance, the total number
of rotor revolutions during the design life is used to define the so-called ‘equivalent 1P
stress range’ where 1P denotes the revolution frequency [7.20].
Likewise the total number of stress cycles, ntotal , during the duration, Tprocess , of the
process with is used sometimes as reference number of cycles [7.10].
T
ntotal ' process ' j nj
(7.5)
Tp
j
For us it is convenient to define the equivalent stress range with respect to the number
of endured cycles, NR, associated with the reference fatigue strength or detail category,
∆σR, of the considered S-N curve, i.e. 2·106 according to EUROCODE or Germanischer
Lloyd (Figure 7.17). 5
5
Another load parameter is useful in the particular case of a bi-linear S-N curve. In a design situation the
structure’s geometry has to be adjusted until the design stress is below the design strength. Therefore the
stress reserve with respect to a unity damage rather than the actual damage is of interest. For this purpose
van Delft at al. [7.21] introduced the so-called K1 - value. For a given stress range distribution the reference
fatigue strength, ∆σR, of a fictitious S-N curve is varied iteratively until the damage approximates one. The
gained reference fatigue strength is denoted as the ‘K1 - value’ where the superscript 1 refers to unity
damage. The K1 - value, also written as ∆σR1, can be regarded as the required reference fatigue strength,
however, this is meaningful only in a design optimisation process and not if a measure for the load
magnitude is required.
The K1 - value, rather than the equivalent stress range, is linear in any load factor of the stress range
distribution. Hence the ratio of reference fatigue strength and K1 - value represents the design’s reserve
in fatigue strength which is also denoted as ‘stress reserve factor’.
Our investigation is restricted to S-N curve with constant slope, so the equivalent stress range, ∆σeq, and
the K1 - value are identical. During a possible future extension on bi-linear S-N curves the K1 - value should
be considered. It provides a kind of natural weighting of the shape of the stress range distribution which
is useful if fatigue loads with different magnitudes are combined.
µ
∆
σeq '
ntotal
NR
4
m p(∆σ ) ∆σ d∆σ .
149
1
µ
j ∆ σj n j ( ∆ σj )
NR j
µ
µ
0
(7.6)
For a bi-linear S-N curve the equivalent stress range is calculated from the inverse S-N
curve ∆σ(N) as function of the damage, D, from (7.4).
∆
σeq '
∆
σ
NR
(7.7)
D
Bandwidth effects
The wide frequency range of the wind loading and partially the different frequency
content of the wind and waves result in a broad banded response of OWEC support
structures. Figures 7.18 and 7.19 illustrate typical spectra and signals with different
bandwidths.
Different parameters are used to measure the bandwidth of a signal. We apply the
regularity factor, β, and the more theoretical bandwidth parameter after Vanmarcke, δ,
[7.16].
m2
Tp
'
β '
(7.8)
Tz
m0 m4
δ
'
1 &
m1
2
(7.9)
m0 m2
For Gaussian processes the regularity factor represents the ratio between peak period
Tp , i.e. the mean time between peaks, and zero up-crossing period Tz , i.e. the mean
time between up-crossings of the mean value. Narrow bandwidth is treated only as the
specific case with β = 1 and δ = 0 of a generally broad banded process characterised
by 0 # β < 1 and 0 < δ # 1.
The probability density function of stress ranges of a narrow banded signal equals the
Rayleigh distribution of the stress peaks and an analytical expression for the fatigue
damage exists [7.22].
However, all positive troughs and negative peaks are ignored and all positive peaks
define ranges of double amplitude regardless whether they actually form stress cycles.
For broad banded signals the approach therefore overestimates the probability of large
stress ranges and so any damage calculation tends to be conservative. Various
researchers have proposed semi-empirical or analytical relations to overcome this
shortcoming, e.g. [7.10], [7.23 - 7.27]. Independent comparisons [7.5, 7.9, 7.11, 7.28],
proved Dirlik’s method (7.10) to be currently the best frequency domain approach for
broad banded processes while Rainflow Counting of time series of sufficient length is
preferred in the time domain.
150
Chapter 7
D1
Q
p(∆σ ) '
&
e
Z
Q
%
D2 Z
R
2
&
e
Z2
2R
2
&
% D3 Z e
Z2
2
Dirlik expression
for distribition of
2 m0
Rainflow ranges
where:
m2
β'
γ'
m0 m4
D1 '
Q'
2
2(γ&β 2)
1%β 2
1.25 ( β & D3 & D2 R )
D1
m1
m2
m0
m4
D2 '
(7.10)
1 & β & D1 % D1
2
1&R
β & γ & D1
R'
D3 ' 1 & D1 & D2
2
1 & β & D1 % D1
2
∆
Z'
σ
2 m0
Weighted quadratic superposition of equivalent stress ranges
Section 7.1 has examined conditions under which the simultaneous response can be
approximated by superposition of separate responses under wind and wave loading.
Hence, the auto spectral density, Sσσ,ah, of the simultaneous response is composed of
the linear superposition of the auto spectral densities, Sσσ,a and Sσσ,h, of the separate
responses and likewise the superposition is valid for the n th spectral moment, mn.
Here index ‘a’ denotes all aerodynamic related loading and response, comprising direct
effects due to wind loading of the structure as well as any indirect effects due to rotor
operation, functional forces and structural response. Likewise, all hydrodynamic effects
are marked by index ‘h’, whilst ‘ah’ refers to the combined response.
4
4
o
o
mn , ah ' m Sσ σ, ah ( f ) f n d f ' m Sσ σ, a ( f ) % Sσ σ, h ( f ) f n d f
(7.11)
' mn , a % mn , h
Note that the spectral moments are defined here as functions of the cyclic frequency,
f, in units of Hertz.6
The superposition of the energy contents, i.e. the variances or zeroth moments, of the
separate responses determines the basis for the superposition of aerodynamic and
hydrodynamic fatigue loads. The peaks of a narrow banded process are Rayleigh
6
Using the relation between the spectral densities in either cyclic or circular frequency S(f) = 2
the spectral moments, mn , expressed in the circular frequency
m n (ω) = ( 2π) n m n(f) , n ≥ 0
to the power of n, times the moments related to cyclic frequency.
π S(ω)
equal 2 π
151
distributed. Applying (7.6) we find a relation between the variance and the equivalent
stress range which can be easily resolved with respect to the variances. 7
µ
∆
σ eq ' 8 m0
ntotal
NR
2%µ
2
Γ
(7.12)
Unfortunately, both the aerodynamic and the simultaneous response are not narrow
banded. Dirlik’s expression (7.10) is capable of such cases but cannot be resolved with
respect to the variance. In the current situation we are looking, however, only for a
formula to link the equivalent stress ranges rather than for the magnitude of the
separate fatigue loads. In this respect the less accurate, empirical correction of the
analytical narrow banded solution after Hancock and Gall, [7.10, 7.24] is more
convenient. We rewrite the original form in terms of the reference number of cycles, NR,
instead of the total number of stress cycles, ntotal, (7.13).
µ
∆
σ eq '
β
8 m0
ntotal
NR
Γ
2%µ
2
Hancock expression
(7.14)
Substitution of the variances in (7.11) leads to:
∆
σ eq , ah '
2
∆ σ eq , a
βah ntotal , ah
2
µ
%
βa ntotal , a
2
∆ σ eq , h
2
µ
βah ntotal , ah
βh ntotal , h
(7.15)
After exchange of the total number of cycles and the regularity factor from Eqn. (7.5)
and (7.8) one obtains a compact relation. The equivalent stress ranges of the two
responses are weighted by the µ th root of the ratio between the zero-crossing periods,
Tz , of the component processes and the combined response. Successively the
contributions are added as the root of the sum of the squares.
∆
σ eq , ah '
2
∆ σ eq , a
T z, a
T z, ah
2
µ
%
2
∆ σ eq , h
T z, h
2
µ
(7.16)
T z, ah
The weighting factors are substituted in terms of the zeroth and second spectral
moment, m0 and m2 , since the zero-crossing period of the combined response, Tz,ah ,
is not directly available.
After this final arrangement we find the desired expression for the so-called ‘weighted
quadratic superposition of equivalent stress ranges’ (7.17).
7
Neglecting the bandwidth one might derive a superposition formula also from the analytical solution under
the rough assumption that the total number of cycles of the simultaneous response equals the sum of the
total number of cycles of the individual responses.
∆
σ eq , ah .
∆
2
σ eq , a
ntotal, a% ntotal, h
ntotal, a
2
µ
2
% ∆ σ eq , h
ntotal, a% ntotal, h
2
µ
(7.13)
ntotal, h
Due to the good results later in this section, the expression was, however, not further investigated.
152
Chapter 7
Weighted quadratic superposition of equivalent stress ranges
µ
∆
σ eq , ah '
m2, a % m2, h
m0, a % m0, h
µ
2
∆ σ eq , a
m 0, a
m 2, a
µ
%
2
∆ σ eq , h
m 0, h
(7.17)
m 2, h
All required quantities are available in the frequency domain or can be estimated with
Eqn. (7.2) in the time domain.
Direct quadratic superposition of equivalent fatigue loads
In Sections 7.2.3 and 7.3 a comprehensive numerical investigation of the validity of Eqn.
(7.17) will be carried out. Prior to this, however, a further simplification of the relation is
proposed from a practical rather than a theoretical point of view.
A more convenient relation is obtained if the weighting factors are omitted, assuming
approximately equal zero-crossing periods of the individual responses. The Pythagoras
law is recovered, which can also be interpreted as 90o out-of-phase superposition of the
equivalent stress ranges (7.18).
Direct quadratic superposition of equivalent stress ranges
∆
σ eq , ah .
∆
2
σ eq , a %
∆
2
σ eq , h
for Tz, a . Tz, h
(7.18)
In most cases the above assumption is not valid in a strict sense because the
aerodynamic response is generally associated with a shorter zero-crossing period than
the hydrodynamic response. Therefore, particular care is required during the following
numerical verification of the simplified expression.
7.2.3 Empirical validation of superposition approaches
Outline of the simulation study
The validity of the two derived formulas for the superposition of the equivalent stress
ranges, Eqn. (7.17) and (7.18), was investigated by an extensive numerical study
including parametric variation of the shape and the bandwidth of the individual response
spectra. In addition, the influence of the ratio of the variances and ratio of the zerocrossing frequencies of the component processes was analysed.
Because of the great bandwidth of some of the combined responses only a comparison
in the time domain was meaningful. The superposition of the equivalent stress ranges
derived from the time series of the separate responses was compared to the result of
Rainflow Counting of the superimposed time series. So, any influence of the methods
described in Section 7.1, as separation of wind and wave contribution and substitution
of aerodynamic damping was not included in this investigation. See Section 8.4 for such
an overall validation.
The complexity of the investigation was increased successively for four different types
of responses.
153
•
block bimodal spectra with δ = 0.1
Guided by Jiao and Moan [7.6] block bimodal spectra were used corresponding to two
narrow banded processes with a Vanmarcke bandwidth, δ, of 0.1. The ratio of variances
ranged from 0.1, 0.3, etc. to 0.9 whilst the zero-crossing frequencies differed by factors
of 0.8, 1.2, 1.6, 3.2 and 7.9.
•
block bimodal spectra with δ1 = 0.1 and δ2 = 0.3
As above, but with distinct Vanmarcke bandwidths.8 Figure 7.18 presents typical power
spectra and associated time series.
•
simplified response spectra
Analytical wind and wave response spectra were derived from a single-degree-offreedom system with an eigenfrequency of 1 Hz and von-Karman turbulence and
Pierson-Moskowitz wave spectra. The damping constant was varied artifically to
construct different bandwidth values and different contributions of the quasi-static and
dynamic part of the response (Figure 7.19). In particular the low-frequency quasi-static
wind response resulted in a broad band. Again the ratio of the component variances
varied from 0.1 to 0.9.
2
T
___
z1
= 0.8
1 Tz2
5
S
2
S
0
1
0
4
−5
5
0
−5
4
5
T
___
z1
= 1.6
2 Tz2
−5
T
___
z1
= 3.2
Tz2
0
20
10
0
0
β = 0.81
0
0
5
β = 0.89
0
stress [MPa]
2
auto spectral density [MPa .s]
T
___
z1
= 1.2
2 Tz2
10
β = 0.92
β = 0.67
0
−5
5
T
___
z1
= 7.9
T
β = 0.62
0
z2
0.5
5
1
−5
0
50
100
150
time [s]
Figure 7.18: Bimodal spectra and corresponding time series for different
ratios between the zero-crossing frequencies of the component processes
(Vanmarcke bandwidth of components δ1 = 0.1 and δ2 = 0.3, variances 0.5 MPa2)
8
For the block spectra the Vanmarcke bandwidths δ1 = 0.1 and δ2 = 0.3 correspond to regularities of
β1 = 0.981 and β2 = 0.87.
154
Chapter 7
•
OWEC response spectra
Finally time series for three load cases of the wind and wave response of the reference
design were evaluated. The ratio between the variances ranged from 0.4 to 13
depending on the considered cross section and load case. Figure 7.20 presents the
different frequency contents of the wind and wave responses for one example. In
distinction to the simplified response spectra, several peaks, corresponding to the foreaft modes and the multiples of the blade passing frequency 2P, are observed in the wind
response spectrum and the bandwidth is further increased. The wave response has
moderate bandwidth and is only significant at the fundamental support structure mode
of 0.29 Hz and the quasi-static wave excitation.
Time series were generated by Inverse Fast Fourier Transform (IFFT) and randomly
chosen phase angles in the first three cases. The DUWECS simulation code was
applied for the OWEC spectra. The minimum number of stress cycles in any response
was 2300 for the block bimodal spectra, 7100 for the simplified and 7700 for the OWEC
response spectra. Generally each stress cycle was represented by at least 10 time
steps. The time series of the OWEC responses were sampled with 13 Hz. For the high
frequency wind response this corresponded on average only to 5.5 points per cycle,
however, in such cases more than 80000 cycles were evaluated.
Discussion of results
At the beginning the overall accuracy of the four methods was studied. The analytical
method after Jiao and Moan [7.6], the evaluation of the superimposed response spectra
with the Dirlik approach (7.10) and the two formulas for the superposition of the
equivalent stress ranges were compared for block bimodal spectra with Vanmarcke
bandwidths of the components of δ1 = 0.1 and δ2 = 0.3.
Figure 7.21 plots the relative error of the equivalent fatigue loading against the regularity
of the combined response ß. Generally the error increased for greater bandwidth, i.e.
lower ß. Only the approach after Jiao and Moan behaved in the opposite manner
because it requires two narrow banded component processes in distinctly different
frequency ranges.
Both the direct quadratic superposition of the equivalent stress ranges (7.18) and the
Dirlik approach exhibit a tendency for slightly nonconservative results for a greater
bandwidth. Larger absolute errors but with positive, i.e. conservative, sign were found
for the weighted quadratic superposition of the equivalent stress ranges (7.17).
Generally the accuracy of the two new formulas was at least as good as any of the
alternatives.
Comparison of the error of the weighted quadratic superposition of the equivalent stress
ranges (7.17) for all four types of response spectra in Figure 7.22 was surprising. The
best approximations were observed for the OWEC spectra rather than for any simplified
spectrum. Typical errors accounted for -3% to 5% of the equivalent fatigue loading,
which is regarded as a quite reasonable accuracy for a simplified approach.
The direct quadratic superposition (7.18) showed a smaller scatter and smaller
magnitude of errors than the weighted quadratic superposition (Figure 7.23). However,
the clear tendency to nonconservative results of up to -5% is unfavourable.
log (auto spectral density) [MPa^2.s]
1E+02
155
1E+02
damping = 100%, ß = 0.26
damping = 25%, ß = 0.67
damping = 25%, ß = 0.58
damping = 0.05%, ß = 0.96
damping = 0.05%, ß = 0.985
1E+01
1E+01
1E+00
1E+00
1E-01
1E-01
1E-02
0.01
1E-02
0.10
1.00
10.000.0
0.2
log (cyclic frequency) [Hz]
0.4
0.6
0.8
1.0
1.2
Figure 7.19: Simplified spectra: Response of a single degree of freedom system to
wind excitation (left) and wave excitation (right)
(Variances normalised to unity.)
log (PSD) [MNm^2.s]
1E+03
1. 00E+03
1E+02
1. 00E+02
1E+01
1. 00E+01
1E+00
1. 00E+00
1E-01
1.00E-01
1E-02
0.01
1.00E-02
0.10
1.00
0.01
10.00
log cyclic frequency [Hz]
wind (ß=0.25)
1P, 2P, 4P, etc.
fore-aft mode
blade flap mode
0.10
1.00
10.00
log (cyclic frequency) [Hz]
wind & wave (ß=0.26)
wave & aerodyn. damping (ß=0.94)
Figure 7.20: OWEC response spectra for bending of soft-soft monopile below
mudline (Vw =14 m/s, Hs = 1.67 m, Tz = 3.34 s):
pure wind response (left), simultaneous wind and wave response (right, solid line),
wave response with aerodynamic damping (right, dashed line)
156
Chapter 7
20%
error of equiv. fatigue load [-]
bimodal spectra:
δ_1 = 0.1, δ_2 = 0.3
direct quadratic
superposition
15%
10%
weighted quadratic
superposition
5%
analytical solution after
Jiao & Moan
0%
-5%
combined spectra &
Dirlik approach
-10%
0.0
0.2
0.4
0.6
ß : regularity [-]
0.8
1.0
Figure 7.21: Relative error of the equivalent fatigue load derived by four
different methods for bimodal response spectra
7.3 Empirical validation of superposition of separate fatigue analyses
7.3.1
Short-term fatigue loads
So far we have learnt to analyse wind and wave responses of an OWEC separately
(Section 7.1) and to combine the resulting equivalent fatigue loads (Section 7.2). The
individual methods showed reasonable accuracy for a simplified approach, however, the
accumulation of errors through the entire trajectory has not been investigated, yet.
For this purpose the equivalent fatigue loads of the reference design at seven hot spots
were analysed for 17 lumped production load cases (Table 8.3). Aerodynamic loading
was calculated by time domain simulations with a commonly used duration of 10
minutes per wind speed class. Spectral fatigue analysis with the Dirlik approach (7.10)
obtained the hydrodynamic fatigue loads.
95% confidence intervals for the estimated fatigue loading were derived by statistical
evaluation of 12 representations of 10 minutes. Per load case the average of 9
simulations of 40 minutes each was used as reference of either pure aerodynamic or
simultaneous wind and wave loading.
In order to avoid bias by application of different models and tools, all analyses except
the operations in the frequency domain were performed with the DUWECS code. For
instance, the transfer functions were derived from time domain simulations of the steady
state wave response while in practice an offshore design tool might establish them
directly in the frequency domain. Likewise the aerodynamic damping was accounted for
in steps of 0.5% after DUWECS linearisation.
Matching of the hydrodynamic model in the frequency domain and the corresponding
model used for the integrated time domain simulation of the simultaneous response was
checked by comparing 20 wave load cases without aerodynamic damping. Figure 7.24
157
error of equiv. fatigue load [-]
20%
bimodal spectra
(δ_1 = δ_2 = 0. 1)
15%
10%
bimodal spectra
(δ_1 = .1, δ_2 = .3)
5%
simplified OWEC
spectra
0%
-5%
real OWEC spectra
-10%
0.0
0.2
0.4
0.6
0.8
1.0
ß : regularity [-]
Figure 7.22: Error of weighted quadratic superposition of equivalent
reference stress ranges (7.17) for different response spectra
20%
bimodal spectra
(δ_1 = δ_2 = 0. 1)
rel. error of equiv. load [-]
15%
10%
bimodal spectra
(δ_1 = .1, δ_2 = .3)
5%
simplified OWEC
spectra
0%
-5%
real OWEC spectra
-10%
0.0
0.2
0.4
0.6
0.8
1.0
ß : regularity [-]
Figure 7.23: Error of direct quadratic superposition of equivalent reference
stress ranges (7.18) for different response spectra
158
Chapter 7
plots the relative difference as function of wave height for the lowest and the highest
hot spot considered. Despite the long simulations with each 9 time series of 40 minutes,
still some scattering is observed due to the random representations. The general
agreement is very good which can be seen from the interpolated trend lines and the
extrapolation to the entire lifetime. The frequency domain approach shows slightly
higher loads especially in the top of the structure.
Four combinations resulting from the magnitude of considered aerodynamic damping,
either 80% or 100%, and of either the weighted or the direct quadratic superposition
were analysed. Figure 7.25 presents the relative error of the combined loading and the
95% confidence intervals against the mean wind speed.
Both methods show similar accuracy. Depending on the cross section and load case the
mean error of an individual load case differs ±10% while the spreading due to only one
simulation of 10 minutes accounts for additional ±3% to ±5%. The direct quadratic
superposition suffers slightly smaller errors but a general conclusion can hardly been
drawn from the short-term behaviour.
7.3.2
Long-term fatigue loads
Comparison of the long-term fatigue loads accumulated from all the load cases on the
right hand side of the diagram shows much better accuracy. Which is shown in Figure
7.26 presenting the long-term error along the entire structure. Accounting for 100% of
aerodynamic damping nearly all results are slightly non-conservative. So the
recommendation from Section 7.1.3 is confirmed to consider only 80% of the estimated
aerodynamic damping as additional structural damping.
The direct quadratic superposition has a spreading of the mean error of about ± 2% in
contrast to ±5% for the weighted quadratic superposition. Likewise the confidence
intervals of ±1.8% around the mean error are smaller in relation to the value of ±2% for
the other approach.
The larger errors of the weighted superposition probably accumulate from the error in
the estimation of the zeroth and especially of the second spectral moment of the
aerodynamic response (7.2). Hofland [7.1] estimated confidence intervals for the
standard deviation between 3% and 8% and for the zero crossing period between 8%
to 10% depending on the simulation length.
Overall evaluation
A benchmark exercise of one particular design proved both formulas for the
superposition of separate fatigue analyses robust and accurate enough for simplified
approaches. In practice, errors are likely to be significantly higher, say in the order of
±10% due to application of different tools for the aerodynamic and hydrodynamic fatigue
analyses, rougher approximation of the aerodynamic damping and execution as an
engineering approach. Therefore a conservative damping assumption should be used.
The relatively low sensitivity of the fatigue loads might be specific to designs with
a high degree of aerodynamic damping such as the pitch regulated reference design.
The weighted quadratic superposition, (7.17), is founded on a more stable basis by
application of the Hancock formula, (7.14), which at least empirically accounts for a
broad bandwidth. The numerical validation showed, however, lower accuracy compared
to the very simple direct quadratic superposition which was derived by ignoring the
considerable bandwidth of the simultaneous wind and wave response. Further
investigations of both formulas for instance with different reference designs should be
carried out. For the time being, the direct quadratic superposition is preferred.
159
ref. difference FD minus TD [-]
3.0%
3.00%
2.0%
2.00%
1.0%
+48 m
(MSL)
1.00%
+48 m (MSL)
0.0%
0.00%
-37m (MSL)
-1.0%
-37m
(MSL)
-1.00%
-2.0%
-2.00%
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
lifetime
significant wave height Hs [m]
Figure 7.24: Relative difference between wave induced equivalent fatigue loads
computed in frequency domain and time domain
Weighted quadratic superpos.: 80% aerodyn. damp. −21 m (MSL)
relative error [%]
20
10
0
−10
−20
6
8
10
12
14
16
18
20
22
Direct quadratic superposition: 80% aerodyn. damp. −21 m (MSL)
relative error [%]
20
10
0
−10
6
8
10
12
14
16
18
20
22
Figure 7.25: Relative error of superposition of short-term equivalent fatigue loads
from separate analyses with respect to integrated time domain simulation of the
simultaneous response
Solid:
Dashed:
Stars:
error when using the average of 9 x 40 min of aerodynamic response
95% confidence interval for using only 1 x 10 min per load case
Individual error from using 1 x 10 min per load case
160
24
0
−12
−21
−29
−37
−7.5 −5 −2.5
80% aerodyn.
damping
0
2.5
5
7.5
48
24
height above MSL [m]
Weighted quadratic superposition
48
100% aerodyn.
damping
Chapter 7
Direct quadratic superposition
48
80% aerodyn.
damping
24
0
−12
−21
−29
−37
−7.5 −5 −2.5
0
2.5
5
7.5
48
24
100% aerodyn.
damping
0
0
−12
−12
−21
−21
−29
−29
−37
−37
−7.5 −5 −2.5 0 2.5 5 7.5
−7.5 −5 −2.5 0 2.5 5 7.5
rel. error of long−term equiv. stress range
rel. [%]
error of long−term equiv. stress range [%]
Figure 7.26: Relative error of long-term equivalent fatigue loads along the structure
(accumulated from superposition of separate short-term fatigue loads) with respect
to integrated time domain simulations of the simultaneous response:
Solid:
Dashed:
Error when using the average of 9 x 40 min simulations per load case
95% confidence interval for using 1 x 10 min simulation per load case
All required quantities of the two expressions are available in the frequency domain or
can be estimated with Eqn. (7.2) in the time domain. This allows application of the
preferred approaches for fatigue analysis in offshore technology and wind energy,
respectively.
In the next chapter the two expressions are implemented the simplified analyses
approaches for the simultaneous wind and wave fatigue covering the entire range from
the environmental and structural modelling to response analysis.
Simplified Analysis of Aerodynamic and Hydrodynamic Fatigue
161
CHAPTER 8
SIMPLIFIED ANALYSIS OF AERODYNAMIC AND HYDRODYNAMIC FATIGUE
The previous chapter has established the theoretical and empirical basis for both the
separation of the simultaneous response of an offshore wind energy converter into an
aerodynamic and hydrodynamic contribution and the successive superposition of the
aerodynamic and hydrodynamic damage equivalent fatigue loads. We have restricted
the discussion to stationary Gaussian processes, representing short-term load cases,
because only under this condition spectral superposition is valid. In the present chapter
the previous findings are applied to propose simplified analysis approaches for the longterm fatigue due to many load cases.
In Sections 8.1 and 8.2 we introduce each one approach for the simplified fatigue
analysis. The difference lies in the superposition based on either the short-term or the
long-term fatigue loads. The treatments fit well to the requirements of the early design
process since they are relatively simple and are based on the application of standard
design tools for the time-domain simulation of the wind response and the linear spectral
analysis of the wave response (Figure 8.1).
For the certification level still integrated, non-linear time domain simulations of the
simultaneous response are preferred, however, the scattering of wind and wave
parameters results in too many load cases from a practical point of view. Therefore
wind contribution
+
•
•
simultaneous
response
wave contribution
non-linear time domain
•
simulation of wind response
wind turbine and support structure •
in a calm sea
•
≈
linear spectral analysis of wave
response
support structure with top mass
additional structural damping
= 80% aerodynamic damping
Figure 8.1: Simplified analysis approach for aerodynamic and hydrodynamic fatigue
of the support structure
162
Chapter 8
Section 8.3, by using the method from Section 8.1, establishes an approach for
constructing a low number of so-called ‘lumped’ load cases (Section 6.3.2) which
approximate the fatigue deterioration due to the complete environmental characteristics.
In addition to the random, stationary load cases, Section 8.3 constitutes also transition
and fault load cases which should be included in a certification calculation. The
simplified approaches introduced in Section 8.1 and 8.2 skip such occurrences because
in most cases their contribution to the fatigue damage is moderate and integrated time
domain analyses are required for instationary situations.
Finally, Section 8.4 compares the three approaches introduced in the previous sections
for a reference example.
8.1 Simplified analysis based on superposition of short-term fatigue loads
A straightforward and well-based solution of the long-term fatigue is to superimpose the
aerodynamic and hydrodynamic fatigue loads associated with the same load case and
to accumulate successively the long-term fatigue from the superimposed short-term
loads.
Figure 8.2 shows the five steps of this approach which are illustrated at the example of
the soft-soft Opti-OWECS monopile configuration described in Section 9.1
Step 1.:
Establishment of environmental conditions
For a simplified analysis it is generally sufficient to consider collinear and omnidirectional
wind and waves. Consequently the environmental conditions can be described by a
three-dimensional scatter diagram with classes for the significant wave height Hs, zerocrossing period Tz and mean wind speed at hub height V. In many situations of an early
design stage no correlated wind and wave data or not even a Hs - Tz scatter diagram of
the site might be available and one has to accept one of the simplified environmental
models introduced in Sections 10.1.1 and 10.2.1.
In the present example a three-dimensional scatter diagram (Table 8.1) is constructed
for classes of ∆Hs = 0.5 m, ∆Tz = 0.5 s and ∆V = 2 m/s by evaluation of the long-term
NESS database [8.1]. From the theoretically possible 990 combinations of the 10 wave
height, 9 wave period and 11 wind speed classes only 87 load cases have a probability
of 0.5 %o and higher.
In the following a distinction between long-term and short-term quantities is introduced
by marking the latter with the indices i, j and k of the corresponding wave height, wave
period and wind speed class.
The consideration of the operational modes of the OWEC is important because they
determine whether or not aerodynamic damping has to be accounted for in the
calculation of the hydrodynamic response associated with winds between the cut-in and
the cut-out wind speed. So in principle two scatter diagrams are required for the
production wind speed range: one with the load cases probabilities during actual
production and another for failure and repair state.
Establishment of such load case probabilities is far from easy. A non-uniform distribution
of the down-time can be expected since the access for maintenance is hindered during
periods with strong winds, high waves or sea ice. Evaluation of Monte-Carlo simulations
of the operation and maintenance behaviour of the entire wind farm [8.2] can derive a
proper estimation. In this chapter we consider a uniform distribution.
163
A conservative treatment must consider the worst result of two analyses with different
assumed availability. The assumption of the minimum expected availability within the
wind farm is necessary for design configurations and hot spots, for which the response
during shut-down, i.e. mainly hydrodynamic loading without aerodynamic damping,
exceeds the combined aerodynamic and hydrodynamic response with aerodynamic
damping during power production at the same environmental conditions. In case of a
dominating aerodynamic loading and for locations closer to the tower top, the maximum
availability, say for convenience 100%, results in the highest loading.
Step 1.
scatter diagram (Hs,i , Tz,j , Vk )
Step 3.a
aerodynamic damping ξ (Vk )
damping classes ξl (Vk)
for each ξ l
hydrodynamic-stress transfer
functions
Step 2.
aerodynamic fatigue analysis
for each Vk
Step 3.b
hydrodynamic fatigue analysis
for each (Hs,i , Tz,j , Vk (ξl ))
stochastic wind field
wave spectrum
non-linear simulation of
OWEC wind response
linear spectral analysis of
support structure response
signal analysis,
Rainflow Counting
spectral moments,
Dirlik range distribution
∆σeq,a,k |p=1 ( m0 , m2 )
∆σeq,h,i,j,k |p=1 ( m0 , m2 )
Step 4.
for each (Hs,i , Tz,j , Vk (ξl ))
superposition of short-term
fatigue loads ∆σeq,ah,I,j,k |p=1
Step 5.
long-term fatigue load ∆σeq,ah
damage Dah
Figure 8.2: Simplified analysis based upon superposition
of short-term equivalent fatigue loads
164
Chapter 8
Table 8.1: Step 1. of simplified analysis: Three dimensional scatter diagram for the
NL-1 site and 100% availability
Cell notation: wind speed class at hub height [m/s] : probability [parts per thousands]
4.75 m
>23: 1
2.5 m # Hs < 3 m
5 s # Tz < 5.5 s
4.25 m
3.75 m
22: 1
>23: 2
13 m/s # V < 15 m/s: p = 0.2%
15 m/s # V < 17 m/s: p = 1.0%
17 m/s # V < 19 m/s: p = 1.3%
19 m/s # V < 21 m/s: p = 0.5%
21 m/s # V < 23 m/s: p = 0.1%
20: 1
22: 2
>23: 1
16:
18:
20:
22:
3.25 m
14: 2
16: 10
18: 13
20: 5
22: 1
2.75 m
2.25 m
14: 8
16: 12
18: 5
20: 1
10: 1
12: 6
14: 17
16: 18
18: 4
20: 1
1.75 m
12: 1
14: 5
16: 2
8: 2
10: 16
12: 43
14: 39
16: 11
18: 1
8: 1
10: 3
12: 3
1.25 m
10: 4
12: 2
6: 4
8: 28
10: 70
12: 57
14: 14
16: 1
<5: 3
6: 10
8: 17
10: 15
12: 5
<5: 1
6: 1
6: 4
8: 18
10: 6
<5: 12
6: 46
8: 76
10: 58
12: 8
<5: 29
6: 40
8: 29
10: 7
<5: 17
6: 9
8: 2
<5: 2
0.75 m
0.25 m
<5: 2
<5: 18
6: 25
8: 8
<5: 34
6: 16
8: 2
<5: 9
6: 2
<5: 1
Hs
2.75 s
3.25 s
3.75 s
4.25 s
4.75 s
Tz
20: 1
22: 1
18: 1
20: 3
22: 2
2
6
7
2
14: 2
16: 4
18: 3
Σ = 986
5.25 s
5.75 s
6.25 s
6.75 s
165
30
60
2nd spectral moment
sectral moment
[MPa^2], [MPa^2/s^2]
25
20
50
equivalent stress range
(design life 20 years,
monopile 5.5 m below mudline)
40
15
30
10
20
5
10
0
equivalent stress range [MPa]
0th spectral moment
0
6
8
10
12
14
16
18
wind speed class [m/s]
20
22
Figure 8.3: Step 2. of simplified analysis: Results of the aerodynamic fatigue
analysis against wind speed class with unit probability
Step 2.:
Aerodynamic fatigue analysis for each wind speed load case
The aerodynamic fatigue response is analysed separately from the hydrodynamics by
means of a standard wind turbine design tool. The aerodynamic response of the OWEC
in a calm sea is computed by non-linear time domain simulations (Figures 8.1 and 8.2,
left). After Rainflow counting of the signals the equivalent stress ranges are determined
according to Eqn. (7.6) for a unity probability of each load case.
Only if one applies the weighted quadratic superposition, Eqn (7.17), the zeroth and the
second spectral moments, m0 and m2, are estimated from the standard deviation and
the counted zero-crossing period with respect to the signal mean by means of Eqn.
(7.2).
Figure 8.3 shows a nearly proportional increase of the equivalent stress range and the
spectral moments against the wind speed. In the vicinity of the rated wind speed some
irregularities occur due to the switching of the pitch controller. Aerodynamic loading
outside the production wind speed range and direct wind loading of the support structure
do not give a significant contribution and are therefore neglected (Section 9.6.1).
Step 3.a:
Determination of hydrodynamic transfer functions
The aerodynamic damping during the operation of the wind turbine is considered as
additional structural damping of the fundamental fore-aft mode of the support structure
during the calculation of the hydrodynamic transfer functions. In Section 7.3 it has been
recommended to reduce the estimated aerodynamic damping by 20% to achieve more
conservative results.
First the dependency of the aerodynamic damping from the wind speed classes, Vk,
established in Step 1., is investigated. Either numerical linearisation or transient
vibration analysis in steady inflow are recommended (Section 7.1.3). Depending on the
166
Chapter 8
observed variation of the damping, which is a consequence of the employed wind
turbine and support structure, damping classes, ξl , are established. At least two classes
are needed for the situation without any aerodynamic damping and with an average
aerodynamic damping. Generally damping classes of 0.5% to 1% of critical damping
provide a more accurate model. For each damping class a specific set of hydrodynamic
transfer functions is derived.
In the example case the aerodynamic damping is rounded in steps of 1% and 0.5% for
values greater to 2% and below 2%, respectively. Successively only 80% of the rounded
values are considered (Table 8.2). By this the 11 wind speed classes are associated
Table 8.2: Step 3.a of simplified analysis: Establishment of damping classes
Wind speed class
Vk [m/s]
< cut-in or
unavailable
6
8
10
12
14
16
18
20
22
> cut-out or
unavailable
index k
Estimated aerodynamic
damping ξ (Vk) [-]
1
Additional structural damping class ξl
ξl [-]
index l
0.0%
0.0%
1
2
3
4
5
6
7
8
9
10
5.2%
5.2%
5.1%
4.3%
3.5%
3.6%
2.2%
1.6%
1.0%
4.0%
4.0%
4.0%
3.2%
3.2%
2.4%
1.6%
1.2%
0.8%
6
6
6
5
5
5
4
3
2
11
0.0%
0.0%
1
200
stress transfer function [MPa/m]
180
monopile 5.5 m below mudline
f o = 0.278 Hz
160
140
0%
120
0.8%
100
1.2%
80
1.6%
60
3.2%
40
4%
quasi-static
20
0
0.00
0.10
0.20
0.30
0.40
0.50
0.60
Figure 8.4: Step 3.a of simplified analysis: Hydrodynamic transfer functions from
wave height to stress range with different values of additional structural damping
(Damping without aerodynamic damping equals 0.5%.)
167
with six damping classes for which the hydrodynamic stress transfer functions are
derived by time domain simulation of the steady state response (Figure 8.4). Depending
on the available tools also a frequency domain analysis is suitable.
Step 3.b:
Hydrodynamic fatigue analysis for each wave load case
The equivalent stress range and if required the zeroth and second spectral moments are
calculated for each load case by linear spectral analysis (Section 6.2.2). Again unit
probabilities are considered. The correct transfer functions are chosen according to the
relation between mean wind speed and damping class from Table 8.2. This treatment
results in the repetition of some identical calculations if for a certain set of Hs and Tz the
same damping class is associated with different wind speed classes. Nonetheless, the
frequency domain approach is so rapid that this hardly affects the computational time.
Figures 8.5 and 8.6 illustrate the entire trajectory of the fatigue analysis process for two
sea sates and different operation conditions at the critical cross section 5.5 m below the
mudline.
The spectral analysis is shown in three left subplots. For the light sea state (Figure 8.5)
the spectral peak of the Pierson-Moskowitz spectrum (top-left) lies close to the
resonance of the transfer function (middle-left) and the response spectrum (bottom-left)
is dominated by the dynamic response.
The right three subplots illustrate the fatigue calculation. The histogram of the stress
ranges (top-right) is computed with the Dirlik formula, (7.10), from the spectral moments
of the response power spectrum. Applying the Palmgren-Miner rule, (7.4), the damage
can be determined from the stress range histogram and the S-N curve. The horizontal
distance between the graph of the histogram and the S-N curve is a qualitative
illustration for the lifetime. Cumulative range distributions (middle-right) are more
commonly used. Apparently only the stress magnitude rather than the number of stress
cycles is affected by the damping. The equivalent stress range (bottom-right), (7.6),
equals a rectangular cumulative distribution owing to the constant range with 2•106
cycles. Without aerodynamic damping the equivalent stress range is 2.8 times higher
corresponding to a 98% shorter lifetime.
The severe sea states in Figure 8.6 includes 19 times more energy which is proportional
to the square of the significant wave height. The lower frequency contents results,
however, in a lower dynamic amplification and a larger contribution of the quasi-static
response especially in case of presence of aerodynamic damping. The difference
between production and parking is much lower than in the previous example because
for this particular wind turbine design only few aerodynamic damping is observed at high
wind speeds.
Step 4.:
Superposition of short-term fatigue loads
The equivalent stress range of the combined fatigue for each load case, ∆σeq,ah,i,j,k , is
composed from the direct quadratic superposition, (7.18), or the weighted quadratic
superposition, (7.17). Application of the direct quadratic superposition is simpler and
probably less scatter of the results occur. The equivalent stress ranges are related to
a unit probability and are marked with a corresponding subscript, p = 1.
168
Chapter 8
Wave spectrum H =0.75 m, T =3.75 s
s
log(range) [MPa]
2
PSD [m *s]
0.2
0.1
0
Stress transfer function
2
10
0
10
Cummulative range distribution (p = 1)
60
range [MPa]
200
[MPa/m]
Histogram of ranges (p = 1), S−N curve
z
0.3
100
40
20
0
0
Response power spectrum
PSDprod
range [MPa]
[103*MPa2*s]
4
Equivalent stress range (p = 1)
60
scale 50:1
2
0
0.1
0.2
0.3
40
20
0 4
10
0.4
5
6
10
7
10
10
log (N)
8
10
Figure 8.5: Spectral fatigue analysis of a light sea state with unity probability
(Dashed: production at 8 m/s with 4% of additional structural damping
Solid: parked with only 0.5% actual structural damping)
Step 5.:
Calculation of long-term fatigue loads
The actual probability of the individual load cases, pi,j,k , is considered during the summation of the long-term aerodynamic and hydrodynamic fatigue response, ∆σeq,ah .
µ
∆σeq, ah '
j j j ∆σeq, ah, i, j, k
µ
i
j
k
p'1
pi, j, k
(8.1)
The total damage from all load cases, Dah , can be calculated from the equivalent stress
range associated with the long-term fatigue, (8.2).
NR
Dah '
(8.2)
N ( ∆σeq, ah )
Alternatively the damages of the individual load cases are summed up, (8.3).
NR pi, j, k
Dah ' j j j
i
j
k
N ( ∆σeq, ah, i, j, k
)
p'1
(8.3)
Wave spectrum H =3.25 m, T =5.75 s
s
log(range) [MPa]
2
PSD [m *s]
Histogram of ranges (p = 1), S−N curve
z
10
5
0
range [MPa]
200
[MPa/m]
2
10
0
10
Cummulative range distribution (p = 1)
100
Stress transfer function
100
0
50
0
Response power spectrum
Equivalent stress range (p = 1)
10
PSDprod scale 50:1
0
0.05 0.1 0.15 0.2 0.25 0.3 0.35
range [MPa]
2
3
[10 *MPa *s]
30
20
169
100
50
0 4
10
5
10
6
7
10
10
log (N)
8
10
Figure 8.6: Spectral fatigue analysis of a severe sea state with unity probability
(Dashed: production at 16 m/s with 2.4% of additional structural damping
Solid: parked with only 0.5% actual structural damping)
In both cases the number of endured cycles N(∆σ) is obtained from the S-N curve, (7.3).
In the example case the hydrodynamic response without aerodynamic damping is more
severe for cross sections below the mudline than the combined aerodynamic and
hydrodynamic response during power production. Consequently the fatigue damage
below the mudline is higher if the OWEC availability is reduced. Figure 8.7 presents the
distribution of the relative damage summed over the considered wind speed classes for
the two cases of 100% and 85% availability. In the former case the damages from the
categories ‘parking below cut-in and above cut-out wind’ and ‘production at 100%
availability’ are summed up. In the latter case a 50% higher total damage is suffered due
to the conditions ‘parking below cut-in and above cut-out wind’, ‘parking at 15% of
production time (85% availability)’ and ‘production at 85% of production time (85%
availability)’.
For most wind speed classes in the production range the damage during the 15% of
time without aerodynamic damping is higher than during the remaining fraction of actual
production. The importance of a proper description of the availability is emphasized.
170
Chapter 8
100
%
16%
ref. damage (all classes) [%]
rel. damage per class [%]
14%
12%
10%
8%
6%
4%
2%
75%
50%
25%
0%
0%
<
cutin
6
8
10 12 14 16 18
wind speed class [m/s]
20
22
>
cutout
all classes
(entire lifetime)
parking below cut-in and above cut-out wind
production at 100 % availability
parking at 15% of production time (85% availability)
production at 85% of production time (85% availability)
Figure 8.7: Step 5. of simplified analysis: Distribution of damage on wind speed
classes for different OWEC availabilities
(Total damage for 85% availability normalised to one.)
8.2 Simplified analysis based on superposition of long-term fatigue loads
In an early design situation and for parametric studies it would be advantageous to use
a further simplified fatigue approach than introduced in the previous section. Empirical
studies have proven that this can be achieved with a reasonable accuracy by the direct
superposition of the long-term equivalent fatigue loads associated with aerodynamic and
hydrodynamic response.
Figure 8.8 explains the four steps. The main difference with the previous approach lies
in the reverse order of the summation of the long-term fatigue, done now separately for
wind and waves in Step 2. and 3.b, and the superposition of aerodynamic and
hydrodynamic fatigue, carried out in Step 4. Consequently, a fifth step is not necessary.
Step 1.: Establishment of environmental conditions
The simplified description of the aerodynamic damping is another distinction. Instead
of several damping classes, now only two situations are considered: the idle, shut-down
or failure state without any aerodynamic damping, hereafter called ‘parking’, and the
production state with assumption of an average value of aerodynamic damping.
Therefore only wave criteria for the parking state rather than a detailed wind-wave
correlation are required. If only separate wind and wave data are available, the cut-in
and cut-out wind speeds can be transformed into corresponding wave heights under the
Step 1.
171
scatter diagram (Hs ,Tz , V)
or
scatter diagram (Hs ,Tz ) & wind speed distribution
load case distribution for chosen OWEC availability
wind speed class distribution
for production or
production and parking
wave load case distribution
for production and parking
Step 3.a
estimate effective aerodynamic damping
during production
for production and parking
hydrodynamic stress transfer
functions
Step 2.
Step 3.b
aerodynamic fatigue analysis
for each wind class
stochastic wind field
hydrodynamic fatigue analysis for all wave
load cases during production & parking
wave spectrum
non-linear simulation of
OWEC wind response
linear spectral analysis of support
structure response
Rainflow Counting
Dirlik range distribution
superposition of stress
range distribution
superposition of stress range
distribution
long-term aerodynamic fatigue
Step 4.
long-term hydrodynamic fatigue
direct quadratic superposition of long-term fatigue loads
Figure 8.8: Simplified analysis based upon superposition of long-term equivalent
fatigue loads by direct quadratic superposition
assumption of a monotonic wind speed - wave height relation. Correlating wind speeds
and wave heights with the same cumulative probabilities can derive such a relation.
As long as it is unknown whether aerodynamic or hydrodynamic response dominates
for all members of the structure one has to evaluate both a minimum and a maximum
availability (see first step of the previous method).
Next the probabilities of the separated aerodynamic and hydrodynamic load cases for
the two operational modes are established by evaluating the Weibull wind speed
distribution and a two-dimensional wave scatter diagram.
Again both a minimum and a maximum availability has to be taken into account for
designs with potentially dominant hydrodynamic response.
172
Chapter 8
Step 2.: Analysis of long-term aerodynamic fatigue
In most cases it is reasonable to consider fatigue only when the rotor is operating in the
production wind speed range. Below the cut-in wind, fatigue is absolutely negligible and
above the cut-out wind speed the hydrodynamic loads are dominating. Therefore we
calculate only the aerodynamic fatigue response during production.
A standard wind turbine design tool is used for the computation of the equivalent stress
range of the aerodynamic fatigue. We do not need the spectral moments because Step
4. applies the direct quadratic superposition of the fatigue loads, (7.18). The equivalent
stress range, ∆σeq,a,prod , is summed from the contributions of the individual load cases,
(8.4).
µ
∆σeq, a , prod '
j ∆σeq, a , prod, k
µ
k
/ p'1
pk
(8.4)
It can also be calculated directly from the long-term stress range distribution, (8.5).
µ
∆
σeq , a , prod '
1
µ
j ∆ σa , prod, j n j ( ∆ σj )
NR j
(8.5)
Step 3.a:
Estimation of the effective aerodynamic damping and determination of
hydrodynamic transfer functions
Estimation of an average or effective aerodynamic damping for the production load
cases is no straightforward task; even if the variation of the damping with the wind
speed is known. Generally sea states with higher waves but relatively low probability of
occurrence contribute significantly to the long-term fatigue damage. Such conditions
correspond to high wind speeds above the rated wind speed for which the aerodynamic
damping might be lower. Figure 8.9 correlates the additional structural damping and the
damage as functions of the wind speed. In this case the damage distribution was
available from the previous section.
Once such a distribution of the hydrodynamic damage on the wind speed classes is
known, at least approximately, a reasonable estimate of the effective aerodynamic
damping ξaero,prod is gained by weighting the aerodynamic damping of the different wind
speed classes ξaero (Vk) by the relative contribution of the damage D (Vk) of the
associated wind speed classes.
j ξ aero ( V k ) D ( V k )
ξ aero , prod . k
/0
(8.6)
00
j D (V k )
0V # V < V
k
in
out
The aerodynamic damping may be estimated from the closed-form expression, Eqn.
(5.12). However, care is required for the influence of the controller dynamics and for stall
regulated design where the aerodynamic drag is significant at higher winds.
Only two sets of hydrodynamic stress transfer functions corresponding to both damping
classes are required.
relative damage [-]
addtional structural damping
[%]
173
4.0%
3.2%
2.4%
1.6%
80% of estimated aerodyn. damping
0.8%
damping class
0.0%
5
7
9
11
13
15
17
19
21
23
25
25%
20%
15%
10%
5%
0%
6
8
10
12
14
16
18
20
22
mean wind speed at hub height [m/s]
Figure 8.9: Establishment of effective damping during production mode:
Additional structural damping against wind speed (upper),
Relative hydrodynamic damage against wind speed (lower)
(Effective additional damping equals 2%.)
Step 3.b:
Analysis of long-term hydrodynamic fatigue
Hydrodynamic fatigue can be significant during production as well as parking. In Step
2. we have assumed that aerodynamic fatigue is negligible during parking. Therefore
the total fatigue for this operating state, ∆σeq,ah,parking , equals the hydrodynamic fatigue
response during parking, ∆σeq,h,parking .
∆σeq, ah, parking :' ∆σeq, h, parking
(8.7)
The long-term equivalent stress ranges for the parking and the production state are
calculated separately from spectral fatigue analyses of the concerned wave load cases.
The probability of the production load cases is adjusted by the assumed maximum or
minimum availability. Either the long-term, equivalent stress ranges are summed from
the individual load cases or they are calculated directly from the long-term stress range
distribution analogous to the aerodynamic fatigue.
Step 4.:
Superposition of aerodynamic and hydrodynamic fatigue
Now we can superimpose the long-term fatigue loading during production by the direct
quadratic superposition of the aerodynamic and hydrodynamic equivalent stress range,
Eqn. (7.18).
2
2
∆σeq, ah, prod ' ∆σeq, a, prod % ∆σeq, h, prod
(8.8)
174
Chapter 8
A straightforward combination of the fatigue during production and during parking would
be the summation of the associated damages.1 Dealing with the damage equivalent
stress ranges from (8.7) and (8.8) we have to exponentiate them with the inverse slope
of the S-N curve before summing up.
∆σeq, ah '
µ
µ
µ
∆σeq, ah, parking % ∆σeq, ah, prod
(8.9)
The total damage can be obtained by (8.2).
8.3 Establishment of load cases for integrated time domain simulations
For an offshore wind energy converter at least an order of magnitude more fatigue load
cases exist than for an onshore wind energy converter. The scatter diagram in Table
8.1, for instance, comprises 87 sea states. Later in this chapter we show that it is more
accurate to deal also with some of the rare but severe load cases which are not included
in this scatter diagram. They rise the total number of sea states to 145 or 235 depending
on the width of the wind speed classes. Consideration of the different operational states
of the turbine results in approximately the double number and taking directional effects
into account would increases the number of load cases by another order of magnitude.
Evidently hundreds of fatigue load cases are incompatible with integrated time domain
simulations and establishment of a low number of characteristic load cases for
certification calculations (Section 10.3) is required.
Step 1.
elementary load cases from scatter diagram
H s ,i , T z , j , V k
Damage calculation of the elementary load cases
Step 2.
weighted quadratic superposition of
aerodynamic and hydrodynamic fatigue,
calculation of damage
Step 3.
Step 4.
preliminary lumping of load cases
Hˆ s( 1,l) , Tˆz(,1l ) , Vˆl (1)
damage calculation of the lumped load cases
ˆ (n)
Hˆ s(n),l , Tˆz(n)
,l , V l
Step 5.
refined lumping of load cases
Step 6.
re-calculation of the damage
and check at different cross sections
Figure 8.10: Lumping of load cases by superposition of short-term fatigue loads
1
In contrast to the discussion on the combination of the fatigue of two simultaneous processes in Section
7.2.1, a summation of damages is valid here because the processes are occurring sequentially.
175
A lumping of load cases is well known in offshore technology and is done most times
under the assumption of quasi-static wave response [8.3]. In the present case the
problem is more involved since dynamic response is important and offshore wind energy
converters suffer both significant wind and wave fatigue. The simplified method
introduced in Section 8.1 offers a possibility for an effective and accurate solution. The
approach has already been mentioned briefly in Section 6.3.2 but is now explained in
full detail (Figure 8.10).
In addition to the lumped load cases, which consider only random, stationary production
and parking conditions, we establish a number of transition and fault situations at the
end of this section.
Step 1.:
Establishment of environmental conditions
A three-dimensional scatter diagram for significant wave height, Hs, zero-crossing
period, Tz, and mean wind speed at hub height, V, is required as starting point. The
corresponding load cases are denoted here as ‘elementary load cases’ and the
parameters, Hs,i , Tz,j , Vk , are marked by indices i, j and k. If directional effects are
significant, the entire procedure should be applied for each directional sector (Section
9.6.1, Table B.2).
Step 2.:
Damage calculation of the elementary load cases
The damage of every elementary load case due to wind and wave loading is determined
with the approach from Section 8.1.
Step 3.:
Preliminary lumping of sea states
Next a first lumping of load cases is carried out. Criteria for lumping of sea states can
be to achieve approximately the same damage for all load cases or lumping of all sea
states associated with a certain wind speed class or a combination of both.
No exact procedure for the lumping process is known. Instead here a practical treatment
is proposed which includes some simplifications and arbitrary elements. The number of
lumped load cases, typically about 20, should not be chosen too small. Good
experience was gained by maintaining most of the wind speed classes with class width
of 1 m/s within the production range but by using broader wind speed classes near the
cut-in and cut-out wind speeds.
The process of lumping itself consists of a weighting of the environmental parameters
of the elementary load cases by their relative damage. Ideally the Hs-Tz-V set of a
lumped load case lies on the centre of damage of the considered elementary load
cases. In fact only the damage at certain values of the environmental parameters rather
than the quite non-linear damage distribution between the discrete parameters is known.
So, an iterative procedure has to be followed. A reasonable first estimate or prediction
step of the lumped parameters can be gained under the simplified assumptions of Eqn.
(8.10), which are derived from the quasi-static response.
D
% ∆σ µ
% Hs
D
% ntotal
%
D
% ∆σ µ
% V
µ
1
Tz
µ
(8.10)
176
Chapter 8
Assuming a linear response, the stress ranges are considered proportional to the
significant wave height as well as the standard deviation of the wind speed and thus
mean wind speed, presuming constant turbulence intensity. The number of stress cycles
is of lower influence on the damage than the stress range. Therefore we suppose the
number of stress cycles to vary with the inverse zero-crossing period which holds, strictly
only for a quasi-static response. The damage itself is proportional to the µth power of the
stress ranges and the total number of cycles, ntotal .
(n)
(n)
(n)
Now the first estimate for the lumped sea state parameters, Ĥs, l , T̂z, l , V̂ l
where
n = 1, is calculated as the centre of damage within the set of the elementary load cases,
Sl. The new index l refers to the number of the lumped load case while the superscript,
n, of the load case parameters denotes the iteration stage.
pi j k
µ
µ
H
p
j
j
j
j j j s,i ijk
i
j
k Tz, j
(1)
1
i
j
k
'
Ĥs, l '
(1)
/0
T̂z, l
j j j pi j k
j j j pi j k /00 (i , j , k ) 0 S
0 (i , j , k ) 0 S
i
j
k
µ
V̂ l
(1)
i
l
'
j j j
i
j
pi j k
k
j
k
l
(8.11)
µ
Vk
j j j pi j k
i
j
k
/0
0 (i , j , k ) 0 S l
Step 4.:
Damage calculation of the lumped load cases
The aerodynamic and hydrodynamic fatigue associated with the lumped load cases is
calculated separately and the combined fatigue is superimposed for each load case.
The required parameters of the aerodynamic fatigue can be interpolated from the results
of Step 2. if the mean wind speed of some load case has been altered. This avoids a
cumbersome new time domain simulations.
Step 5.:
Refined lumping of load cases
The damage of the lumped load cases will not match exactly the summed damage
associated with the elementary load cases owing to the simplifications in Eqn. (8.10).
The corrections, (8.12) and (8.13), of the lumped significant wave height and inverse
zero-crossing period with the same factor, vD, (8.14), should be suitable if the wind
speed of the load cases is not modified.
(n % 1)
Ĥ s
(n)
' νD Ĥ s
(8.12)
(n)
(n % 1)
T̂ z
'
T̂ z
νD
(8.13)
The correction, vD, is also based upon the simplified, quasi-static relations from Eqn.
(8.10). Approximately the µ/(µ+1)th fraction of the error of the damage of the lumped load
177
cases is corrected by changing the significant wave height while the remaining 1/(µ+1)th
part of the error is compensated by the variation of the zero crossing period.2
µ%1
νD, l '
j j j D Hs, i , Tz, j , Vk
i
j
k
D
(n)
Ĥ s ,
(n)
T̂ z
, V̂
(n)
(i , j , k ) 0 S l
(8.14)
In some cases with dominant dynamic response owing to a zero-crossing period,
(n)
T̂z, l
, in the vicinity of the fundamental eigenfrequency, the iteration may not
converge. Then only the significant wave height should be corrected by (8.12).
Consequently vD is calculated by applying the µth root in (8.14).
Step 6.:
Re-calculation of combined fatigue and check for different cross sections
Similar to Step 4., the combined fatigue is re-computed for the refined lumped load
cases. The error of the damage should be checked at cross sections with different ratio
between the aerodynamic and the hydrodynamic fatigue. The procedure is repeated
from Step 3. or Step 5. if the deviation between the damage associated with the
elementary load cases and the lumped load cases is not acceptable.
Example of load case lumping
The NESS database enables us to consider a minimum probability of 0.50/000 and
classes of ∆Hs = 0.5 m, ∆Tz = 0.5 s and ∆V = 1 m/s which results in 235 sea states. This
number enlarges to 440 elementary load cases when parking of the turbine within the
production wind speed range is taken into account. As illustration Figure 6.7 presents
the scattering of all 3-hour sea states included in the NESS database for a period of 9
years.
By following Step 2. to 6. we can established 21 lumped sea states according to Table
8.3. From Figure 6.7 is can be seen that these load cases are covering very well the
spreading of the actual conditions. If a reduced availability is considered, the load cases
within the production wind speed range, i.e. No. 3 to 19, have to be simulated for
production as well as parking and a total number of 37 load cases exists.
Table B.2 in Appendix B presents lumped load cases for directional sectors with a width
of 30o. Opposite sectors are combined, so, only six sectors must be evaluated.
The approximately 10 times lower number of integrated time domain simulations, which
remain after the sea state lumping, reduces the computational time accordingly. The
effort for the entire lumping procedure depends mainly on the manual operations
conveniently done with a spreadsheet during Step 1., 3. and 5. These operations may
take one hour in total. The damage calculations in Step 2., 4. and 6. require only some
minutes if the aerodynamic fatigue loads are provided and tools for linear spectral
analysis and the superposition of fatigue loads exist.
Establishment of transition and fault load cases
So far we have only considered fatigue due to stationary, random aerodynamic and
hydrodynamic loading during normal production or parking. Fatigue is also caused by
2
Assuming an inverse slope of µ = 4 and an error of 10% then a correction of both Hs and Tz by
approximately 2%, i.e. νD = 1.019, is sufficient. The damage is adjusted by a factor of 1.0194 = 1.079 due
to the modification of Hs and by a factor of 1.019 owing to the new zero-crossing period.
178
Chapter 8
Table 8.3: Lumped load cases for the Opti-OWECS design
at the NL-1 site for 100% availability
No.
Hs [m]
Tz [s]
V1h [m/s]
IT [%]
Probability [o/oo]
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
0.48
0.80
0.72
0.83
0.99
1.10
1.27
1.45
1.62
1.83
2.03
2.23
2.47
2.68
2.89
3.15
3.40
3.64
3.84
3.97
5.19
3.57
4.50
3.88
3.88
4.16
4.23
4.39
4.55
4.68
4.82
4.94
5.05
5.22
5.34
5.48
5.66
5.84
5.97
6.11
6.03
6.98
<5
<5
6.1
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18.5
19.5
20.5
21.5
22.5
> 23 †
> 23 †
none
none
11.8
11.4
11.2
10.9
10.7
10.6
10.4
10.2
9.9
9.6
9.4
9.2
9.1
9.0
8.9
8.8
8.7
none
none
66
63
158
88
95
90
91
71
58
47
41
35
25
21
14
11
9
6
4
4
2
Sum
999
V1h = hourly mean wind speed at 60 m height
IT = turbulence intensity for a turbine spacing of 10 diameters (Figure 4.7)
†
The assumed shut-down condition for the WTS 80M is either a 3 s gust of 30 m/s or
higher or a 10 minute mean wind speed greater than 25 m/s. From a statistical point of
view both cases correspond to a hourly mean wind speed of 23 m/s at hub height if
hysteresis effects are accounted for. If turbulent wind loads are considered during parking
one may assume for these load cases a mean wind speed of 0.7·Vref in accordance with
IEC 61400-1 ed.2.
Table 8.4: Transition load cases for the Opti-OWECS design at the NL-1 site
Load case
Hs [m]
Tz [s] V10min [m/s]
Probability
Start-up & shut-down at cut-in wind
0.66
3.44
5
1000 times/year
Start-up & shut-down at cut-out wind
4.79
6.38
25
50 times/year
179
transition between operating conditions, e.g. start-up, shut-down, switching between the
rotor speeds of double speed designs and passing of the tower resonance at soft-soft
design with variable rotor speed. Likewise fault conditions could be of importance which
do not directly initiate a shut-down of the turbine. A typical example includes a
malfunction of the wind vane and associated longer operation with a significant
misalignment between the wind direction and the rotor axis. Manoeuvres with short
duration as emergency shut-down, extreme gusts or wind shears are commonly only
considered in the ultimate limit state rather than fatigue analysis [8.4].
Establishment of any of such load cases depends on the particular design. The sea
state parameters can be derived from the mean conditions taken from the NESS
database. Germanischer Lloyd recommends in its OWECS guidelines [8.5] to take into
account start-up and shut-down procedure with a frequency of 1000 times per year at
cut-in wind and 50 times per year at cut-out wind both combined with a corresponding
sea state (Table 8.4). Section 9.6.1 demonstrate only a minor contribution of these load
cases to the support structure fatigue.
The situation of power production with a large yaw misalignment is very unlikely for the
WTS 80M turbine which is equipped with two independent wind vanes and an extensive
remote monitoring and control system. Therefore the load case is not further considered
here.
8.4 Parameter study
Some practical recommendations and an impression on the accuracy of the three
proposed approaches are gained by a parameter study for soft-soft Opti-OWECS
monopile configuration. In accordance with Section 9.6.1 the critical cross section of the
monopile 5.5 m below mudline and a location 10.6 m underneath of tower top flange are
analysed. As reference we consider the long-term fatigue for 85% availability which is
extrapolated from integrated, non-linear time domain simulations of 440 elementary load
cases with a duration of 40 minutes each. Again a inverse slope of the S-N curve of 4
is applied.
Figure 8.11 presents the error of the damage equivalent stress ranges from the shortterm superposition of fatigue from Section 8.1 for four cases:
•
Short-term, direct quadratic superposition of 160 load cases
corresponding to a minimum load case probability of 0.50/00 and classes of ∆Hs =
0.5 m, ∆Tz = 0.5 s and ∆V = 2 m/s (see scatter diagram in Table 8.1). The total
probability account for 98.6% of the conditions within the NESS database.
•
as above but the minimum probability equals 0.50/000 which covers 99.9% of all sea
states included in the database
•
Short-term, weighted quadratic superposition of the 260 load cases
•
corresponding to a minimum load case probability of 0.50/000 and classes of ∆Hs =
0.5 m, ∆Tz = 0.5 s and ∆V = 1 m/s
180
Chapter 8
Short-term, direct quadatic superpos.:
160 load cases
Short-term, direct quadratic superpos.:
260 load cases
Short-term weighted quadratic
superpos.: 260 load cases
Short-term direct quadratic superpos.:
440 load cases
-6%
Tower (z = 45 m)
-4%
-2%
0%
2%
4%
6%
error of equiv. stress range [%]
Monopile (z = -26.5m)
Figure 8.11: Error of equivalent stress range from short-term superposition
of fatigue loads (Section 8.1) for 85% availability with respect to
integrated time domain simulation of 440 elementary load cases
Long-term superpos.: 1.6% add.
damping
Long-term superpos.: 2% add.
damping
Long-term superpos.: 2.4% add.
damping
Tower (z = 45 m)
-6%
-4%
-2%
0%
2%
4%
6%
Figure 8.12: Error of equivalent stress range from long-term superposition
of fatigue loads (Section 8.2) for 85% availability with respect to
integrated time domain simulation of 440 elementary load cases
Integrated Time Domain simulation of
38 lumped load cases
Tower (z = 45 m)
-6%
-4%
-2%
0%
2%
4%
Figure 8.13: Error of equivalent stress range from integrated time domain
simulation of 21+17 lumped load cases (Section 8.3, Table 8.3)
for 85% availability with respect to 440 elementary load cases
6%
181
In any case of this particular example the accuracy of the simplified method is good
since the error is below 5%. Fatigue of the monopile is underestimated while the
opposite holds for the tower section. The influence of the class width for the wind speed
is less significant than that of the minimum load case probability. If possible with respect
to the available environmental data a minimum load case probability of 0.50/000 is
recommended while a wind speed bin of 2 m/s seems to be sufficient. The difference
between the direct quadratic superposition, (7.18), and the weighted quadratic
superposition, (7.17), is small and the preference for the simpler and more robust direct
quadratic superposition, stated in Section 7.3.2, is confirmed.
The influence of the estimated effective aerodynamic damping during production is
studied for the long-term superposition approach from Section 8.2. Again the errors of
this even simpler method are small and range between -6% and -1%. Now at both cross
sections a too low estimate of the equivalent stress range is observed. The reference
design is quite insensitive with respect to a ±20% variation of the assumed effective
aerodynamic damping which is a typical accuracy of the damping estimation. In this
example the equivalent stress range is affected only by ±2%.
Finally the load case lumping from Table 8.3 in Section 8.3 is verified. Very good
agreement of ± 0.6% with the respect to the result from 12 times more time domain
simulations is found in Figure 8.13. This observation is amazing because the method
rests upon the simplified approach from Section 8.1 which itself is less accurate (Figure
8.11). The most likely explanation is that the simplified approach is used only to weight
the relative contribution of the different elementary load cases to a certain lumped load
case.
The results of the parameter study are promising and clearly prove the suitability of the
new methods for engineering applications during the different stages of the design
process.
In the daily design practice larger errors are expected due to application of different
computational tools and less consistent models for both the environment and the
mechanical system. Consideration of only one particular reference design limits,
however, a general judgement of the accuracy and further investigation of different
offshore wind energy converters are recommended.
182
Chapter 8
183
Interrelation of Dynamics and Design
CHAPTER 9
INTERRELATION OF DYNAMICS AND DESIGN
The dynamic behaviour of an offshore wind energy converter is influenced by the design
properties and site conditions. Vice-versa, the design depends (among other aspects)
on the dynamics. In this chapter the interrelation is investigated for a broad spectrum of
concepts and conditions.
After introducing the reference designs (Figure 9.1, Table 9.1), categorized by the
applied wind turbine concept and size in Section 9.1, five essential aspects of dynamics
and design of offshore wind energy converters (OWEC) are discussed in Section 9.2
to 9.6. Successively we are dealing with the siting and the environmental conditions, the
NL-5
Lely
NL-1
Blekinge
Utgrunden
Figure 9.1: Location of the six reference sites
Table 9.1: Combination of wind turbines, support structures and sites
Wind turbine
Support structure
Site
NedWind 40 †
soft-stiff monopile A3 †
soft-soft monopile A2 †
soft-soft monopile A2-mod ‡
IJsselmeer: Lely A3
IJsselmeer: Lely A2
IJsselmeer: Lely A2
Enron Wind 1.5o †
soft-soft monopile †
Baltic: Utgrunden
stiff-stiff lattice tower ‡
soft-stiff gravity monotower ‡
soft-soft monopile ‡
North Sea: NL-5
Baltic: Blekinge
North Sea: NL-1
Kvaerner Turbin WTS 80 †
WTS 80 †
WTS 80M ‡
†
actually built, ‡ proposed
184
Chapter 9
dynamic characteristics, the aerodynamic and other sources of damping, the foundation
uncertainty and the response behaviour under fatigue and extreme conditions.
It is beyond the scope and not necessary to analyse every item for every design and
site. So only those designs are considered which are suitable to stress important effects
and to derive design recommendations. Section 9.7 provides a more general overview
by comparing the design drivers and complementary results for the different support
structures and sites.
9.1 Reference designs and sites
Overview
A variety of proposed or actually realised designs and sites is chosen for investigation.
Looking at the wind turbine we see three different ratings of 500 kW, 1.5 MW and 3 MW
which reflect the rapid development of offshore wind energy. The 500/600 kW class was
applied for the first demonstration plants in Denmark, the Netherlands and Sweden
between 1991 and 1998. Beginning in the year 2000 the megawatt class is employed
for the first large-scale offshore wind farms. Even larger machines in the multi-megawatt
class have been proposed for very large wind farms on a medium time scale.
Each turbine is of a different design concept being the two-bladed, active-stall controlled,
500 kW NedWind 40, the three-bladed variable-pitch, variable-speed Enron Wind 1.5
offshore and the two-bladed, pitch regulated Kværner Turbin WTS 80 / WTS80M with
constant speed. In fact, both two-bladed machines are representing uncommon design
concepts. Especially the combination of pitch regulation and constant speed of the
machines of Kværner Turbin is aged and not representative for machines in the multimegawatt class which are currently under development.
The NedWind 40 design was realized at the first Dutch offshore wind farm Lely in 1994.
Although it is an example for the parallel design approach (Section 3.3.2); important
lessons about the sensitivity and response of monopile concepts were gained [9.1, 9.2].
Experiences that could be used in a more integral manner during the design and
installation of the Utgrunden offshore wind farm with the Enron Wind 1.5 offshore in the
year 2000. A prototype of the WTS 80, known as ‘Näsudden II’, is operating since the
beginning of the 1990's on the island of Gotland. The sister machine with variable speed
‘Aeolus II’ was erected in Wilhelmshaven, Germany.
The diversity of designs is completed by the consideration of three generic support
structure types, monopile, gravity monotower and gravity lattice tower. Particular
emphasis is given to the monopile configuration. It is investigated for each of the three
wind turbine classes because of its superior economic performance for many sites and
turbine types. Another, not directly visible, difference lies in the dynamic characteristics
which cover the entire span of opportunities from soft-soft, soft-stiff to stiff-stiff.
Sub-megawatt class of the first Dutch offshore wind farm Lely
The first Dutch offshore wind farm Lely (see front cover) comprises four NedWind 40
turbines at the western side of the IJsselmeer. Installed in 1994, steel monopiles were
chosen for the very first time as OWEC foundations.
Figure 9.2 shows the three investigated configurations, denoted Lely A3, A2 and A2mod. The first two are based on the data specified in the original design calculations
[9.3] and some updated data [9.4]. The design A3 has soft-stiff characteristics thanks
185
onshore
A3
A2
A2mod
actual pile penetration
fictitious clamping point
Figure 9.2: Comparison of an onshore design with the
three offshore designs of the wind farm Lely
to a pile with a large diameter of 3.7 m and a long penetration depth of 20.5 m at 5 m
water depth. Soft-soft characteristics were proposed for design A2 at a greater water
depth of 10 m. So a lighter pile of 3.2 m diameter and only 13.5 m penetration was
chosen. Eventually measurements proved soft-stiff characteristics, most likely been
caused by unexpected soil conditions or a modification of the design after the design
calculation. For us, however, investigation of the original design configuration is more
interesting.
By contrast to its two companions, the design A2-mod only exists within this thesis as
an example for a design improvement of case A2 (Section 9.5.3). Although both
concepts show the same fundamental eigenfrequency, the design A2-mod uses a larger
pile penetration depth, a reduction in pile diameter close to the water surface and a
stretched tower.
Megawatt class at the Utgrunden wind farm
With commissioning in autumn 2000 the wind farm on the Utgrunden reef in the
Kalmarsund between the Swedish mainland and the Øland Island was one of the first
projects which employed wind turbines of the megawatt class (Figures 1.2 and 2.5).
Located 12.5 km from the mainland and 8 km from the island, the water depth ranges
between 7.2 and 10 m.
The project with 7 units of the Enron Wind 1.5 offshore type was a turn-key project
developed by Enron Wind. This enabled an integrated design of the wind farm and a first
direct application of results of this thesis.
Utgrunden is also the first project with soft-soft monopile foundations. The foundation
and tower were designed together as one structure, considering loads and frequencies
186
Chapter 9
from machine/rotor, wind, waves, ice, etc,
into an efficient and optimal structure. This
could use a lighter and thus cheaper pile
stop ring and tower. The relatively small pile
tower
diameter of 3 m reduces the installation
efforts and results in lower ice loads. By
flange
transition
this no ice cone was required which itself is
piece
sensitive to ice damage and which cannot
grout
be applied at sites with larger variations in
the sea level e.g. due to tides and storm
surges.
Monopile foundations are very interesting
from an economical point of view. However,
monopile
attention is needed for the inclination of
driven monopiles after installation and the
Figure 9.3: Illustration of grouted joint possible damage of the pile top owing to
the hammer impact. Therefore an
innovative grouted joint between the pile and the tower is used (Figure 9.3). It proved
easy and fast installation and capabilities to correct pile inclinations and compensate
imperfections of the pile top.
In this thesis the location T4 with the maximum water depth of 10 m is considered as a
reference example. The monopiles were designed for a large range of soil conditions
which were difficult to estimate for this site on a glacial moraine with complex
geotechnical constitution. We assume soil parameters in the middle of the design range
which correspond to measurements of the natural frequencies. The fundamental
eigenfrequency of the seven monopiles of approximately 0.27 to 0.29 Hz lies in the
operational range of the mean rotor speed from 0.183 to 0.33 Hz. Therefore for each
OWEC an individual operational window of the rotor speed is excluded from stationary
operation. Above rated wind speed the turbines run at a mean rotor frequency of 0.33
Hz.
Multi-megawatt concepts of the Opti-OWECS
During the Opti-OWECS project three distinct OWEC concepts all using a proposed 3
MW wind turbine design of Kvaerner Turbin AB were subject to detailed dynamic
analyses (Figure 9.4) [9.5]:
• Lattice tower - NL-5 site
A gravity, steel lattice tower with stiff-stiff characteristics is combined with the WTS
80 wind turbine at a site 50 km off the Dutch North Sea coast The stiff framework
with three legs is based on experience in the design of offshore structures for
demanding environments.
• Gravity monotower - Blekinge site
A steel monotower on a gravity base with soft-stiff characteristics carries the WTS
80 wind turbine at a Swedish Baltic site. The considerable ice loading justifies a
robust, single column structure.
• Monopile - NL-1 site
The final design solution of the project includes a steel monopile with soft-soft
characteristics and the WTS 80 M wind turbine at a Dutch North Sea site
40 m
4.5m
14.1 m
58.6 m
Scour Protection
3500 mm Ø
x 75 mm
Boat
MSL Landing
Grouted
Joint
Tower Flange
2800 mm Ø
Tower
J-Tube
Figure 9.4: Opti-OWECS support structures: Left: Gravity lattice tower with three legs at 25 m water depth (MSL),
Middle: Gravity monotower at 15 m water depth (MSL), Right: Steel monopile at 21 m water depth (MSL)
21 m
187
Transition Piece
15 m
Pile (45 m long,
25m penetration)
188
Chapter 9
approximately 15 km offshore. This concept was developed with an integrated
design approach considering the techno-economics of the entire wind farm (Section
3.6, Chapter 10).
The WTS 80 M (marine) machine with a 16% higher rotor speed is the offshore variant
of the WTS 80 with a two-bladed 80 m pitch-regulated, constant-speed rotor, developed
originally for onshore siting [9.6].
The oscillating motion of the flexible, hydraulic yaw system is used to lower the inherent
high, dynamic loads of two-bladed rotor with a rigid hub and constant rotor speed. In
addition, damping is introduced into the yaw motion and other vibration modes of the
system. Within the scope of the thesis it was not possible to accurately model the
characteristics of the yaw system, partly due to restrictions of the used computational
tools and partly due to incomplete design specifications. The assumption of a rigid yaw
system is, however, likely to be conservative. A proper representation would have
required a cumbersome new derivation of the equations of motion in the rotor module
and an advanced model for unsteady aerodynamics and yawed flow. In the modelling
of the control system some simplifications have been made which were not necessary
if more detailed wind turbine data was available.
Further details of the designs and sites are presented in Appendix B.
9.2 Siting
Site characteristics
The six reference sites (Figure 9.1) are representing a broad band of environmental
conditions ranging with increasing water depth and exposure from the two places in the
sheltered inland sea of the IJsselmeer and two ice loaded locations in the Baltic to the
two North Sea sites. The growing severity of the sites is typical for the evolution of
offshore wind energy over time. While sheltered waters were chosen for the
demonstration plants about a decade ago, hostile North Sea locations will probably be
exploited on a medium time scale.
Table B.1 in Appendix B provides the main characteristics of the sites.
Offshore versus onshore wind climate
In Section 4.1.1. we have discussed the differences of the offshore wind climate to the
situation on land, being higher average and extreme wind speeds but lower turbulence,
wind shear and gust factors. Now we will use the soft-soft Opti-OWECS monopile design
to quantify the fatigue effect of different site characteristics.
Figure 9.5 plots contour lines for the same equivalent stress range at the critical cross
section of the sub-soil pile for different annual average wind speeds (x-axis) and design
turbulence intensities at 15 m/s mean wind speed (y-axis). For illustration we have
marked the offshore site NL-1 (Vave = 9 m/s, I15 = 9%) and two onshore site classes
defined by the IEC 61400-1 ed. 2 standard. The IEC TC III-A (type class III: Vave = 7.5
m/s, turbulence class A: I15 = 18%) represents a site with moderate wind speed but high
turbulence while IEC TC II-B could be a windy coastal location with low turbulence (Vave
= 8.5 m/s, turbulence class B: I15 = 16%). The aerodynamic fatigue loading is normalised
by the value for the NL-1 site.
Obviously the fatigue loading is driven by both annual average wind speed and
turbulence intensity, however, the latter has the stronger influence. The equivalent
189
+
0.18
onshore
IEC TC III−A
+ 160 %
onshore
IEC TC II−B
0.16
turbulence intensity I15 [−]
180 %
0.14
120 %
140 %
100 %
0.12
80 %
0.1
0.09
6
offshore
NL−1
+
7
8
9
annual average wind speed [m/s]
10
Figure 9.5: Contour plot of the normalised equivalent stress range at the monopile of
the Opti-OWECS design due to pure wind loading as function of the annual average
wind speed (x-axis) and the turbulence intensity at 15 m/s (y-axis)
(Loading at the NL-1 site equals 100%.)
fatigue loading at the land-based site is approximately 40% to 50% higher than at the
NL-1 location. If the same high turbulence would be assumed at both onshore sites, a
23% higher loading could be expected at the more windy location. In fact, the increase
account only for half of this, thanks to the lower assumed turbulence class.
Large offshore wind farms will take relatively less profit from the low (natural) ambient
turbulence since the wakes of upstream turbines are more persistent and induce a
background turbulence level. For the Opti-OWECS design this effect is not so significant
since a simpler and older wake model (Figure 4.7) and a wide turbine spacing of 10 rotor
diameters is chosen. In practice, half of the distance is more realistic.
Likewise to the turbulence, the relative gust amplitude at extreme wind conditions is
lower offshore. Measured evidence is however scarce. So, standards use different
safety philosophies here. If one follows the Germanischer Lloyd Guidelines the gust wind
speed offshore is not necessarily higher than on land because offshore the extreme gust
is assumed to be only 20% higher than the extreme mean wind speed in comparison to
an increase of 40% onshore (Table 4.2). IEC 61400-1 ed. 2 has been developed for
onshore applications only and uses the same gust factor than Germanischer Lloyd
onshore while DS 472 applies the identical gust factors onshore and offshore.
For pitch regulated wind turbines the extreme operating gust is often an important
extreme load case. Again different models exist in the various standards. Germanischer
190
Chapter 9
9
NL-5 (North Sea, NL)
omnidirectional significant
wave height [m]
8
NL-1 (North Sea, NL)
7
Blekinge (Baltic Sea, SE)
6
Utgrunden (Baltic Sea, SE)
5
Lely (IJsselmeer, NL)
4
3
2
1
0
0
5
10
15
20
25
30
mean wind speed at 10 m height [m/s]
Figure 9.6: Relation between mean wind speed and omnidirectional significant
wave height at the reference sites
(Cut-off limit at extreme 50 year significant wave height)
Lloyd and DS 472 apply site independent gust magnitudes while the IEC 61400-1 ed.2
relates the amplitude to the turbulence intensity and rotor size.
High average wind speeds over 8.5 m/s classify many offshore sites in the demanding
Type Class I according to IEC 61400-1. However, if the lower turbulence is considered
machines which were originally designed for Type Class II might still be accepted by a
site specific certification. Likewise the extreme mean wind speeds of most of such Class
I sites is considerably lower than the 50 m/s assumed by the international standard.
Hydrodynamic and ice conditions
The hydrodynamic conditions vary largely between the sites (Figure 9.6). In addition, the
greater water depth at exposed sites further increases the overturning moment. Roughly
wave loads with double magnitude occur at the NL-5 site compared to the Baltic sites.
The fatigue loading depends also on the dynamics of the structure and the wave
steepness. Dynamic loading of short waves with low height can be significant even in
the IJsselmeer if the design is sensitive and excited close to resonance.
At many Baltic sites the governing extreme load takes place as a combination of the
extreme sea ice and an aerodynamic loading.
9.3 Dynamic characteristics
Analysis of the eigenfrequencies and eigenmodes can provide important, more
qualitative insight in the dynamic behaviour which itself can only be quantified by more
cumbersome response analyses.
191
In this section we give an overview on the natural frequencies of the reference designs
and see how far they are affected by the offshore foundation (Section 9.3.1). Simple
one-degree-of-freedom models can explain the influence of the eigenfrequencies on the
aerodynamic excitation of the tower top and the hydrodynamic forcing of both the
support structure and flapwise blade bending (Section 9.3.2). Finally, in Section 9.3.3,
the relevance of the mode shapes for the sensitivity against variations of the soil
stiffness and the effectiveness of the wave excitation is investigated.
Table 9.2: Eigenfrequencies of the reference designs in Hertz
Design
Lely
fR =0.54
fb =1.08
Eigenmode
A3
A2
A2mod
1st fore-aft
2nd fore-aft
0.70
3.03
0.44
1.63
0.44
2.00
1st blade flap
1st blade lead-lag
2.17
3.26
Utgrunden
variable
speed †
0.28
1.46
WTS 80
fR =0.32
fb =0.63
Lattice
tower
Monotower
0.76
2.69
0.51
1.27
1.03
1.83
1.40
2.11
WTS80M
fR =0.37
fb =0.73
Monopile
0.29
1.29
1.41
2.11
fR = rotor frequency, fb = Nb · fR blade passing frequency
fR = 0.183 - 0.33 Hz and fb = 0.55 -1 Hz for operational range of mean rotor speed
Ω = 11 - 20 rpm, mean rotor speed above rated wind speed 20 rpm
†
9.3.1
Natural frequencies and foundation influence
Natural frequencies
The major eigenfrequencies of the reference designs and the rotor excitation
frequencies are listed in Table 9.2. The blade frequencies are not significantly affected
by the support structure stiffness, thus, no difference is made between the Lely and
WTS 80 designs.
Foundation influence and fixity length of monopiles
Obviously, the design of offshore wind energy converters has to bypass resonances with
the rotor frequency or its multiples and excessive wave excitation. Different
considerations exists for the foundation design compared to land-based buildings or
offshore platforms.1
1
The foundation design of most land-based buildings is often governed by the bearing capacity under
limit state conditions. Stiffness or other dynamic criteria generally are of lower importance.
For onshore wind energy converters both the bearing capacity and the avoidance of a rotor or blade
resonance is important. The latter requirement is often matched by tailoring the tower because the
foundation flexibility reduces the fundamental eigenfrequency only up to 5 to 10% compared to a rigid
foundation. Thus the design is fit for purpose for a variety of sites if the foundation stiffness is located
somewhere between the minimum value assumed and infinity.
The design of most foundations of offshore platforms is governed by limit state loads as well. The
foundation influence on the first natural frequency is between 0% and almost 20 to 40% at small offshore
structures. This considerable spread is caused by the foundation type, the variable soil properties and
(continued...)
192
Chapter 9
This problem is accentuated by the comparison of the foundation influence between the
seven reference designs in Table 9.3. The soil flexibility decreases the fundamental
eigenfrequencies of the two gravity designs, lattice tower and monotower, by about 7%.
In contrast, the monopiles are subject to an at least two times higher reduction. The Lely
A2 case shows an exceptionally high effect of 40%.
Another, more illustrative measure for the flexibility of a piled foundation is the apparent
fixity length introduced in Section 5.2.2. Table 9.3 presents this quantity in units of the
pile diameter. An appreciably low variation between 3.1 and 3.7 is discovered between
the values derived from the Lely measurements and the design values of the Utgrunden
and Opti-OWECS design. For a preliminary design an indicative figure between 3 and
4 pile diameters can be recommended.
The illustration of the fictitious clamping points at the fixity depth of the Lely designs in
Figure 9.2 is useful to explain the dynamic behaviour. Whilst the A3 and A2-mod
monopiles are regarded to be fixed at about 3/4 of the penetration, the fixity point of the
A2 design is located at about the double penetration depth. This seems to be strange
but is confirmed when considering the almost rigid body shape of the first eigenmode
in Figure 9.10.
The sensitivity of the eigenfrequencies is further studied in Section 9.5.2 and Section
10.3.2 where the effect of uncertain soil properties and other influences is studied.
In general, the foundation influence is so large that individual foundation design is
necessary for each wind farm site and each OWEC type. Within a wind farm site,
modifications of the support structure design including the foundation might be required
owing to variation of the water depth and possibly the local soil conditions.
Table 9.3: Quantification of foundation influence
Lely
decrease of 1st
eigenfrequency
compared to
rigid foundation
A3
A2
A2
-mod
14%
13%†
40%
17%†
16%
10%‡
Utgrunden
11%
Opti-OWECS
Lattice
tower
Monotower
Mono pile
7.4%
6.3%
13%
apparent fixity
4.1
9
4.9
length
3.1
not applicable
†
†
‡
3.4
3.4
3.2
[pile diameter]
†
derived from measurements, ‡ extrapolated from Lely A2 measurements
1
3.7
(...continued)
the non-linear soil behaviour under different load levels. This and other considerations favour relatively
stiff offshore structures with natural frequencies well beyond the wave excitation range extending between
f = 0.04 Hz and f = 0.5 to 1.0 Hz. Dynamic wave loading is strongly reduced for structural frequencies
above approximately 0.4 Hz.
193
9.3.2
Excitation of tower and flapwise blade bending
From previous investigations it is well-known that the rotor blades of a (bottom-mounted)
OWEC are affected predominantly by aerodynamic loads whilst the support structure is
subject to both significant aerodynamic and hydrodynamic excitation [9.7, 9.8]. One may
argue that this is an obvious fact since waves may never touch the rotor whereas the
wind and wave loads are transferred though the structure into the soil. But does this also
hold for very soft designs, as the Opti-OWECS monopile, where considerable vibration
amplitudes occur at the tower top?
Application of two models each with a single degree of freedom (SDOF) for the
translation of the tower top and for the flapwise blade motion can explain the basic
response characteristics (Figure 9.7).
Both aerodynamic and hydrodynamic loads are directly exciting the support structure
which corresponds to an excitation of the SDOF system at the spring. The response is
first quasi-static, i.e. the dynamic amplification factor (DAF) is close to one. With
increasing forcing frequency it passes high dynamic amplifications in the resonance
region and decreases finally for high excitation frequencies below the static response.
Figure 9.9 plots the DAF for two damping values associated with the response of a pitch
regulated design in the wind direction, where an aerodynamic damping ratio of 3% is
active, and perpendicular to the wind direction with only 0.5% material damping. The
excitation frequency is made non-dimensional by the fundamental eigenfrequency.
Most important for the aerodynamic response are the rotor frequency (1P) and blade
passing frequency (Nb@P) which are shown with markers for the three Opti-OWECS
designs. The stiff-stiff lattice tower design exhibits a DAF of 3.7 of the strong 2P
excitation whilst the soft-stiff monotower and soft-soft monopile show lower
amplifications of 1.8 and 0.2, respectively. This explains qualitatively, what we will see
during the aerodynamic and hydrodynamic fatigue analyses in Sections 9.6.1 and
10.2.2. For the two-bladed WTS 80 turbine it is quite beneficial to have a soft support
structure even if higher dynamic wave loading has to be accounted for.
Due to the non-dimensional frequency scale the wave excitation range has different
extensions for the three concepts. Nearly quasi-static response occurs for the stiff-stiff
lattice tower. The natural frequency of the soft-stiff monotower lies on the edge of the
wave energy band, however, the soft-soft monopile is located in the middle of it.
Wave excitation of the rotor blades can only take place through inertia forces generated
by the tower top acceleration which is most significant at the structural
eigenfrequencies.2 Here the SDOF system is forced through a prescribed motion of the
frame. The dynamic amplification in Figure 9.8 has a reverse dependency on the nondimensional excitation frequency ftower/fflap given by the ratio between support structure
and flap eigenfrequency [9.9]. The blade follows the low-frequency excitation and does
not feel any significant inertia loads. Beyond the resonance peak the excitation is too
fast and the blade rests more and more in an inertial reference frame.
If we take the Opti-OWECS monopile for an example, the ratio of the first tower and
blade eigenfrequency accounts for 0.2 governing a dynamic amplification of only 0.04.
2
The ratio between blade mass and equivalent tower top mass is so small that for this purpose the
interaction between the support structure with nacelle and the rotor is described as a forced vibration of
the rotor due to a kinematic excitation at the hub.
194
Chapter 9
uflap
u&&top
u&&top
Fwind
Fwave
Fwind
utop
Fwave
dynamic amplification factor DAF [-]
Figure 9.7: Single degree-of-freedom models for the support
structure excitation and flapwise blade excitation
5
separated flow
(1% material damping)
4
excitation by 2nd
tower eigenfreq.
3
2
attached flow
(20% aerodynamic
+ 1% material damping)
excitation by 1st
tower eigenfreq.
1
0
0
0.5
1
1.5
2
2.5
tower eigenfrequency / 1st flap eigenfrequency [-]
wave excitation range (0.04 - 0.5 Hz)
dynamic amplification factor DAF
(log-scale) [-]
Figure 9.8: Dynamic amplification of flapwise blade motion
due to tower top acceleration with frequency equal to
the tower eigenfrequencies
100.0
excitation lateral to
wind direction
excitation in wind direction
(aerodynamically damped)
10.0
stiff-stiff: 2P
soft-soft: 1P
soft-stiff: 2P
soft-stiff: 1P
1.0
soft-soft: 2P
stiff-stiff: 1P
0.1
0
0.5
1
1.5
2
2.5
excitation frequency / fundamental eigenfrequency [-]
stiff-stiff
soft-stiff
soft-soft
wave excitation
range (0.04 - 0.5 Hz)
Figure 9.9: Dynamic amplification of the support structure of the
Opti-OWECS concepts due to aerodynamic and hydrodynamic loads
195
Although a DAF of 1.6 is related to the second support structure eigenfrequency with
ftower/fflap = 0.85 the wave excitation energy is much lower for this mode than for the
fundamental mode. So both modes do not result in a significant blade response either
due to a low DAF or low excitation energy.
9.3.3
Relevance of mode shapes
A visual inspection of the shape of the eigenmodes can be used to judge the
foundation’s sensitivity to the variable soil conditions and to the hydrodynamic excitation.
As example the modes of the three Lely monopile designs and the gravity monotower
are plotted in Figure 9.10.
Foundation sensitivity
Comparison of the subsoil displacements of the Lely designs reveals distinct behaviour.
Designs A3 and A2-mod are properly clamped in the soil whilst design A2 undergoes
large deflection in this area. The first eigenmode of Lely A2 behaves similar to a rigid
body mode, having minor curvature along the structure but large rotational motion in the
soil. Significant lateral displacements occur over the entire penetration depth. In
contrast, only the upper quarter of the soil in case of the first mode and the upper half
of the penetration for the higher modes are affected for Lely A3 and especially A2-mod.
A large contribution of the foundation translation to the entire mode is also observed for
the second mode of the gravity monotower.
Obviously, modes with considerable soil deformation are more sensitive to the soil
conditions. In Section 9.5.2 (Figure 9.13) this is confirmed also quantitatively by a
numerical analysis, which proves high sensitivity in particular for the first modes of Lely
A2 and the second mode of the gravity monotower.
Hydrodynamic sensitivity
In the previous Section 9.3.2 we have studied the dynamic amplification as the result of
the ratio between excitation frequency and natural frequency and of the excitation type.
The distribution of the excitation forces, as wave or wind loads, in relation to the mode
shapes is of importance, as well. For instance, a certain mode of a rod can barely be
excited in the vicinity of a vibration node even if the forcing frequency is close to the
resonance frequency. This effect can be quantified by the magnitude of the generalised
external forces, ΦT·f, on the right hand side of the equations of motion, (6.2). The modal
transformation corresponds to the spatial integration of the product of the concerned
mode shape and the external load.
We are mainly interested in the wave excitation of the support structure and introduce
a hydrodynamic participation factor, χhyd,i , of the i th mode shape, ni (z). Normalising the
maximum displacement of the mode shape to unity and division by the base shear
obtains a non-dimensional expression for the generalized force, (9.1), which has only
a relative meaning.3
3
For the fundamental mode, which has the maximum deflection at the tower top, the hydrodynamic
participation factor can be interpreted as the relative effectiveness of the wave excitation compared to
the forcing by the rotor thrust.
196
Chapter 9
0
m n i (z) dF (z) dz
χhyd, i '
&d
(9.1)
0
max n i (z)
m dF (z) dz
&d
The wave force per unit length, dF(z), shows an exponential decay from the
instantaneous water surface down to the sea bottom. So, small waves are only effective
near the water surface while larger waves (in shallow waters) affect the entire
submerged structure.4
Table 9.4 compares the hydrodynamic participation factors for the six monopile and
monotower designs.5 For the Lely A2 design a two to three times higher hydrodynamic
excitation can be expected than for the other Lely and Opti-OWECS designs if the same
fundamental eigenfrequency is assumed. The Utgrunden monopile has an extraordinary
low participation factor due to the rather stiff monopile in comparison to the slender,
double conical tower. Therefore, mainly the tower flexibility contributes to the first two
bending modes.
Table 9.4: Hydrodynamic participation factors of eigenmodes
Bending
mode
1st mode
2nd mode
3rd mode
9.4
Lely
A3
A2
A2-mod
Utgrunden
5.9 %
57 %
14 %
20 %
80 %
5%
9.1 %
69 %
5%
2.9 %
22 %
25 %
Opti-OWECS
Monotower
Monopile
5.3 %
88 %
27 %
9.2 %
57 %
29 %
Aerodynamic and other sources of damping and inertia
9.4.1 Aerodynamic damping
The most significant source of damping of an offshore wind energy converter is the
aerodynamic damping of the flapwise blade bending modes and the fundamental
support structure fore-aft mode. Other modes show none or much lower damping. In
Section 5.3 we discussed the principles of this effect and we found two major influences
by analysing the simple closed-form expression (5.12): The local derivative of the lift
coefficient with respect to the angle of incidence and the structural dynamics of the wind
turbine and the support structure. Both aspects can be studied at the reference designs.
Section 7.1.3 and Figures 7.10 to 7.12 addressed the aerodynamic damping ratio as
fraction of critical damping of the fundamental fore-aft mode for the pitch regulated WTS
4
In fact the influence of the wave height and wave length is moderate and it is sufficient to apply linear
wave theory which yields an integration between z = 0 at the still water line and z = - d at depth, d, on the
sea bottom. We assume a deterministic wave with the parameters of the most probable sea state.
5
The participation factor, (9.1), is not defined for space frame structures where horizontal and vertical
wave forces are acting simultaneously.
197
50.
Lely A2 50.
40.
40.
30.
30.
20.
10.
SWL
0.
height [m]
height [m]
Lely A3
20.
10.
SWL
0.
mudline -10.
-20.
-10.
mudline
-30.
displacement [-]
50
Lely
A2-mod 40
-30.
displacement [-]
50.
Gravity
40.
Monotower
10
SWL
0
mudline
30.
height [m]
height [m]
30
20
-20.
SWL
20.
10.
0.
-10
-10.
-20
mudline -20.
-30
displacement [-]
-30.
displacement [-]
Figure 9.10: Bending mode shapes of the three Lely designs
and the Opti-OWECS gravity monotower
1st mode,
2nd mode,
3rd mode
aerodynamic damping
(as fraction of critical damping) [%]
198
Chapter 9
2.0%
Lely A2-mod ( fo = 0.44 Hz )
Lely A2 ( fo = 0.44 Hz )
1.5%
Lely A3 ( fo = 0.70 Hz )
1.0%
0.5%
0.0%
5.0
7.5
10.0
12.5
15.0
17.5
20.0
22.5
25.0
wind speed at hub [m/s]
Figure 9.11: Aerodynamic damping of the fundamental fore-aft mode
of the Lely designs
80M wind turbine on a soft-soft monopile. The behaviour for an active-stall regulated
design can be studied at the Lely examples in Figure 9.11.
The difference between stall and pitch regulation is obvious. In both cases the high
damping at low wind speeds reduces towards the rated wind speed. For a stall regulated
design the flow separation reduces the damping ratio dramatically in the vicinity of the
rated wind speed. The increasing drag loading leads to some recovery of the damping
on a low level in deep stall close to the cut-out wind speed. Where in case of pitch
control it recovers beyond the rated wind speed since the reduction of the angle of
incidence results again in more attached flow (Figure 7.11). The fast dynamics of the
pitch control system influence the damping. In this particular case the damping ratio
decreases towards the cut-out wind speed.
Softer support structures with a lower modal mass experience larger aerodynamic
damping ratios. The lower natural frequency of the Lely A2 and A2-mod configurations
result in a higher damping ratio than for the A3 design. The more slender tower of A2mod features a lower modal mass and affect more damping compared to Lely A2 with
the first same eigenfequency. Similar behaviour is observed at the Opti-OWECS designs
where the monopile is about 50% better damped than the gravity monotower and lattice
tower. In the latter case the low modal mass of the stiff framework results in an
appreciably high damping ratio for such a stiff structure of approximately 3% at low wind
speeds where the monotower shows about 3.5%.
In summary, the variation of the damping ratio between different designs and different
operational conditions are so large, that proper modelling is certainly required. Time
domain simulations with an integrated OWEC model consider the aerodynamic damping
inherently. For other approaches, e.g. in the frequency domain, a set of models covering
the full range of operational conditions and associated aerodynamic damping values is
required.
199
9.4.2 Hydrodynamic mass and damping
An effect of the fluid loading of offshore structures, present also in a calm sea, is the
mass of the internal water and the effective mass of the surrounding fluid, (5.11).
Neglecting both of them, would result in an increase of the fundamental eigenfrequency
of the reference designs of less than 2%. Only the Lely A2 design shows a greater
sensitivity of 5%. By contrast, the second bending frequency is significantly affected for
all designs. While the effect accounts for 15% for the stiff lattice tower, it is up to 60%
for the soft Lely design A2. The larger influence can be explained qualitatively by the
greater value of the hydrodynamic participation factor of the second eigenmode in
comparison to the first mode in Table 9.4.
Two general reasons make hydrodynamic damping and vortex shedding not significant
for bottom-mounted OWEC support structures. The submerged part of such structures
is relatively stiff in comparison to the dry section and structural velocities of the wet
members are low. So the hydrodynamic lift and drag forces, which depend quadratically
on the velocities, are also low and any viscous hydrodynamic damping is negligible.
Consequently, the hydrodynamic drag forces are calculated based upon the Morison
equation written in absolute water particle velocities.
There will be a very small amount of damping owing to the radiation of energy from the
submerged surface, which is however not further considered.
9.4.3 Soil damping
Sections 5.2.2. and 5.2.3 described distinct models for both the hysteretic or plastic
damping, more significant for piled foundations and the radiation damping, important for
gravity foundations. It has been attempted to provide damping estimates for both
foundation types.
Analysis of the hysteretic soil damping of the piled foundation is more complex due to
the influence of the lateral loading and applied soil model.
Table 9.5 compares the modal damping ratios of the bending modes of the Lely A2 and
the Opti-OWECS monopile derived from an equivalent viscous damping (5.5). Due to
the restrictions of the mathematical model the values itself are more qualitative and the
most important conclusion is that the fundamental modes show little soil damping while
the higher modes experience damping ratios between 0.15% and 0.6%.
The lateral loading is relatively small compared to the ultimate bearing capacity. So
almost elastic rather than elasto-plastic soil behaviour occurs, which results in low
damping. Furthermore, it should be noted that the prediction of hysteretic soil damping
is strongly dependent on the P-y curve used and the modelling of the soil behaviour
Table 9.5: Estimates of the soil damping of two monopile support structures
Lely A2
Bending
mode No.
Eigenfrequency
Opti-OWECS monopile
Damping
ratio
Bending
mode No.
1 (fore-aft)
. 0.26 %
1 (fore-aft) 0.444 Hz
1 (lateral)
. 0.23 %
0.455 Hz
1 (lateral)
2-6
1.6 - 14 Hz 0.15 - 0.5 %
2-4
(Level of cyclic loading corresponds to operation during
annual average wind speed.)
Eigenfrequency
0.288 Hz
0.292 Hz
1.2 - 15 Hz
the sea state
Damping
ratio
. 0.07 %
. 0.04 %
0.4 - 0.6 %
associated with the
200
Chapter 9
100.0%
damping ratio [%]
2nd torsion
2nd bending
& translation
3rd bending
& rocking
10.0%
4th bending
1.0%
1st torsion
1st bending
0.1%
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
eigenfrequency [Hz]
frequency independent
frequency dependent
Figure 9.12: Eigenfrequencies and soil damping for the gravity monotower
calculated with either frequency independent or frequency dependent soil stiffness
and damping
under low strains. Small differences in the mathematical modelling can have a large
effect.
Indication on the total structural and soil damping can be gained from measurements
during shut-down and employment of the mechanical brake at the Lely A2 and A3
turbines. From the evaluation of the time series of the lateral tower top acceleration with
(7.1) we find damping ratios of 0.85% and 0.7% for both designs (Table 9.6). These
values are larger than the range of 0.2 to 0.5% measured at common onshore wind
turbine towers (Section 5.2.1) [9.10]. For the time being the 0.5% damping ratio used in
the numerical simulations seems to be an adequate number and further investigations
at other monopile structures are recommended before this might been altered.
Closed-form solution for the soil stiffness, (5.6), and soil damping, (5.7) and (5.8), of
gravity bases on an elastic half-space were discussed in Section 5.2.3. For the gravity
monotower the eigenfrequencies and associated damping ratios of a frequency invariant
model are compared to those from the iterative solution of a frequency dependent
modelling in Figure 9.12 [9.11]. In the relevant frequency range up to 5 Hz the natural
frequencies are not affected by the model choice. Small effect is seen on the damping
except for the important fundamental mode where the frequency dependent approach
obtains 0.4% damping which is only half of the value from the simpler method.
The frequency dependent model is only compatible with a harmonic analysis rather than
with transient time domain simulations which is our intention. Thus, neglecting of the
frequency dependency and a conservative assumption of the calculated soil damping
is recommended for design purposes.
Further experimental investigations of both piled and gravity support structure types are
recommended to reduce the uncertainty about the magnitude of soil damping.
9.5
201
Foundation uncertainty
Even after in-situ tests the soil properties are inherently uncertain and soil dynamics of
offshore structures remain a difficult affair. Despite the various mathematical models
mentioned in Sections 5.2.2 and 5.2.3 the most rational treatment is to consider a much
larger safety margin than for the environmental loading and material strength. Load
calculations should consider the most onerous soil conditions which will be in most
cases the lower assumed bound for soil stiffness and strength (Section 10.3.2).
The particularities of OWEC foundations are highlighted by evaluation of eigenfrequency
measurements at the wind farm Lely with an unexpected large deviation from the
proposed design for the Lely A2 converter (Section 9.5.1).
Taking this experience we analyse the foundation sensitivity for the reference designs
and are able to derive some design recommendations in Section 9.5.2 which are used
in Section 9.5.3 to propose a fictitious design improvement for the problematic Lely A2
case.
9.5.1 Evaluation measurements at the Lely wind farm
The Lely project was the first of its kind employing monopile foundations. Therefore
measurements of the eigenfrequencies and the shut-down behaviour of the A2 and A3
designs were carried out in early 1995, six months after installation [9.12].
Both measurements, listed in Table 9.6, confirmed stiffer response behaviour than
predicted by the design calculation [9.3]. For the A3 location the prediction by the
original design calculations was approximately 6% too low. Despite the mismatch this
is no problem for a safe plant operation. An even larger margin to both the rotor
excitation at 0.54 Hz and the wave excitation is provided.
By contrast, the error of 35% at Lely A2 can neither be accepted nor explained
completely. Only the fact that the deviation is large enough to change the characteristics
from the proposed soft-soft to the actual soft-stiff type prevents considerable rotor
excitations and likely fatigue problems. As a side effect the significant wave excitation
of the proposed design is reduced.
A parameter study was performed to explain possible reasons for the discrepancy
between predictions and measurements [9.4]. The design calculations were repeated
with more realistic assumptions, which include consideration of water mass, a lower
lateral prestress level of the piles, slightly revised design data for nacelle mass, tower
wall thickness and mass of the tower interior. The P-y curves of the soil were
maintained, since the derivation from Cone Penetration Tests (CPT) and partly
laboratory tests could be reproduced [9.13]. The errors were lowered by this exercise
(Table 9.6) but do not resolve the wrongdoing behaviour.
So far only speculative explanations can be given. Firstly, the site conditions could be
different than assumed. A lower water depth owing to back-filling of the local sink or
stiffer soil might be the case. Indeed the design is extraordinarily sensitive to the soil
conditions, as will be concluded below. No single reason is sufficient since, for instance,
the soil stiffness must be increased by a factor of 10 (!) to fit the measurement.
Secondly, the design might not correspond to the specifications, e.g. the pile penetration
or pile stiffness are larger or the wind turbine mass is lower. Further queries did not
answer this question.
Thirdly, the measurement at A2 but not at A3 might be wrong or the similar design A1
at a lower water depth was measured instead.
202
Chapter 9
To clarify the situation a repetition of the measurements has been initiated which also
should study possible aging effects [9.14]. At the time of writing the investigation was
ongoing.
Table 9.6: Measured versus predicted first eigenfrequencies at Lely
Lely
†
Measurement
eigenfreq.
damping
ratio †
A2
0.63 Hz
.0.85%
A3
0.72 Hz
. 0.7%
Original design
calculation
eigenfreq.
Calculation with
revised OWEC data
error
eigenfreq.
error
eigenfreq.
on rigid
foundation
0.41
Hz
- 35%
0.44 Hz
- 30%
0.74 Hz
0.68
Hz
- 5.6%
0.70 Hz
-2.8%
0.81 Hz
total structural and soil damping measured during shut-down
9.5.2 Sensitivity of eigenfrequencies
Sensitivity study for the reference designs
Design calculations have to account for the uncertainty of the soil parameters and
models. In Chapter 10 distinct approaches are introduced that fit the requirements of the
design process.
One of them proposed for the conceptual design phase is applied here to study the
sensitivity of the reference designs. The soil stiffness is multiplied by a partial safety
factor γm . A priori it is not clear whether a lower or a higher resistance is more onerous.
Existing design guidelines, for instance Germanischer Lloyd or American Petroleum
Institute provide only safety factors on the soil strength ranging between 1.5 and 2.
Without more specific knowledge about the particular conditions, we may apply
indicative values of γm = 0.5 and γm = 3.0.
The upper tolerance is greater than the reciprocal lower tolerance because soil
parameters are often estimated at the lower side, which is conservative from a strength
point of view. Moreover, the fundamental eigenfrequency is depending decreasingly on
the soil stiffness, so the asymmetrical stiffness variation corresponds to an
approximately symmetrical frequency variation.
Figure 9.13 compares the sensitivities of the first two fore-aft natural frequencies of the
reference designs.
The fundamental eigenfrequency of all designs with the exception of Lely A2 is affected
by ±5% or less which is reasonable. The two gravity foundations show larger sensitivity
than the monopiles. This holds particularly for the comparison with the Utgrunden and
Opti-OWECS monopiles. Lely A2 suffers unacceptabe vulnerable and unsafe behaviour.
The risk of even more extensive wave excitation or a rotor resonance is enormous.
In all cases the sensitivity of the second bending mode is greater than for the first mode.
Again and even more pronounced the gravity foundations perform less favourable than
203
Second eigenfrequency
First eigenfrequency
Lely A3
Lely A2
Lely A2-mod
Utgrunden
Lattice tower
Monotower
Monopile
-25% -15% -5%
5%
15%
1/2 * design stiffness
25% -30%
-10%
10%
30%
50%
3 * design stiffness
Figure 9.13: Sensitivity of first (left) and second (right) fore-aft natural frequency of
the reference designs
the monopiles. Although a resonance of the fundamental mode is more critical, the
second mode could easily interfere with higher harmonics of the rotor frequency or a
blade bending mode.
For completeness it is mentioned that the possibility of scour (Section 4.3) induces
another soil uncertainty that increases the lever arm of the environmental forces and can
lower the eigenfrequencies significantly.
‘Tuning’ of eigenfrequencies
Sometimes it is proposed to ‘tune’ the fundamental eigenfrequency by placing ballast
near the tower top. However, one should be aware of the considerable weight required
and the practical implications of the installation. For instance, 20% extra nacelle weight
decreases the fundamental frequencies of the reference designs between 6.5% and
7.5% only. The second bending mode is hardly affected since the associated translation
of the tower top is small.
Wind turbines with variable rotor speed as the Enron Wind 1.5 offshore at Utgrunden
can tackle eigenfrequencies located ‘somewhere’ within the rotor speed range by
intelligent control and exclusion of an operational window. Nonetheless, difficulties occur
if the eigenfrequency lies in the vicinity of the nominal rotor speed which is commonly
used for operation between the rated and the cut-out wind speed. Alternation of the
nominal rotor speed would have a larger impact on the control system and the
aerodynamic efficiency.
9.5.3 Improved design for wind farm Lely
A too low penetration depth of the Lely A2 design can be concluded from the evaluation
of the eigenmodes in Section 9.3.3 and the sensitivity study in the previous section.
204
Chapter 9
Figure 9.14 illustrates the effect of the penetration depth on the nominal first
eigenfrequency and its indicative sensitivity range for the three Lely designs. When the
penetration depth of A2 is increased beyond that of Lely A3, the frequency tolerance
becomes acceptable but a rotor resonance occurs.
At the A2-mod design the pile penetration is increased from 13.5 m to 20 m, a level
beyond which no further significant increase of stiffness can be achieved and the
uncertainty range of the first eigenfrequency is reduced to ±5%. The soft-soft behaviour
is maintained by employing a pile extension of 1.5 m above the sea level and a 8.5 m
longer tower. The shift of the flexibility from the foundation to the taller tower also
reduces the sensitivity to the soil stiffness compared to both A2 and A3.
0.8
0.7
Design A3
0.6
forbidden area:
0.5
rotor excitation 1P +/- 10%
Design
A2-mod
0.4
0.3
Design A2
0.2
12
14
16
18
20
pile penetration depth [m]
22
24
Figure 9.14: First eigenfrequency of the three Lely designs and uncertainty
ranges as function of the pile penetration depth
9.6
Fatigue and extreme response
9.6.1 Aerodynamic and hydrodynamic fatigue
Different sections of the thesis discuss aerodynamic and hydrodynamic fatigue.
Therefore only some complementary aspects are dealt with here. We address again the
soft-soft monopile of the Opti-OWECS study since it is most challenging in this respect.
In contrast to other sections here only the method of the integrated, non-linear time
domain simulation is applied.
The design is subject to considerable aerodynamic fatigue owing its size and the
concept of a two-bladed rotor with constant speed and pitch regulation. With the
fundamental eigenfrequency in the range of significant wave excitation the concept
suffers also high dynamic amplification of the loading and considerable hydrodynamic
fatigue. In Section 10.2.2 (Figures 10.8 and 10.9) it is demonstrated how the overall
loading has been tailored by a higher rotor speed and a softer support structure.
Start-up and shut-down behaviour
Table 8.4 in Section 8.3 has established load cases for the transition between the
production and parking state and vise versa. In many cases such load conditions are of
205
[rpm]
[deg]
85
20
0
22
16.7
8.35
Pitch angle for start−up at 5m/s stationary wind
Rotor speed (8.35rpm = 2P resonance, 16.7rpm = 1P reson., 22rpm = rated)
[kW]
Generator power
50
25
0
50
25
0
−25
10
[MNm]
[MNm]
[kN]
Axial and lateral rotor force
Fore−aft bending moment My (−5.5m mudline)
5
0
5
Lateral bending moment Mx (−5.5m mudline)
0
−5
0
20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400
time [s]
Figure 9.15: Start-up procedure of the soft-soft monopile configuration and passing
of the blade resonance at 8.35 rpm and rotor resonance at 16.7 rpm
lower importance for the total fatigue loading of the support structure but they are
demanded for certification calculations. The WTS 80M with the soft-soft monopile is an
interesting configuration where during start-up and shut-down the rotor and the blade
excitation frequencies have to pass the fundamental support structure eigenfrequency.
Figure 9.15 illustrates the start-up behaviour at the cut-in wind speed. If a critical wind
speed or rotor idling speed is detected by the supervisory control system the blades are
pitched into the rotor plane (1st signal). In the example a start from the resting position
at 85o is assumed but idling at a low rotor speed with a pitch angle of approximately 60o 65o is possible as well. During the pitching action the rotor is accelerated
aerodynamically (2nd signal) and the pitch rate is reduced. In case the acceleration is not
fast enough to reach the rated speed or the rotor speed remains too long in the vicinity
of the resonance frequency the start-up manoeuvre is cancelled and the blades are
feathered again. In the present case with stationary wind the operation is successful and
the generator is connected after synchronisation to the grid (3rd signal). Owing to the
negligible aerodynamic damping of the lateral motion the two resonances are clearly
visible in the lateral bending moment (6th signal). Tower shadow, wind shear, yawed flow
and nacelle tilt (Figures 5.2 and 5.3) cause the blade passing (2P) excitation of the
structure. The aerodynamic damping leads to a much lower excitation of the fore-aft
response (5th signal) than for the lateral motion (6th signal). The rotor resonance (1P) is
206
Chapter 9
triggered by the mass and aerodynamic imbalances of the rotor. Now both fore-aft and
lateral response are significant. For such a fast passing of the resonance ranges the
vibration amplitudes at the tower top stay below 8 cm and the associated stress
variations are low.
During shut-down at low wind speeds a short employment of the mechanical brake with
a brake torque of 40% of the rated torque is required to pass the blade resonance
quickly enough. Pure aerodynamic braking results in a large overshoot of the tower top
motion into the wind direction since the rotor acts as a propeller when the pitch angle is
too large with respect to the rotor speed.
Starting and stopping at high wind speeds have to be combined with the power control
but enable easy crossing the resonances. The power in the wind is large and enables
quick up-speeding and slowing-down of the rotor in combination with only moderate
pitch angle variations.
The transition load cases from Table 8.4 have been analysed in both directions and at
the cut-in and cut-out wind speeds for stationary wind combined with stochastic waves.
A total damage below 0.003 occurs for pure wind loading. For combined aerodynamic
and hydrodynamic loading the damage in the monopile is still below 0.01 and it is
dominated by the wave response at the cut-in wind speed. In principle such situations
are covered also by the parking load cases No. 2 and 20 in Table 8.3. The probability
of occurrence of these stationary, stochastic conditions has been estimated
conservatively by assuming a high cut-in and low cut-out criterion. It is therefore
concluded that for the present OWEC configuration the transition load cases do not
contribute significantly to the fatigue loading.
Dynamic amplification of wave response
DAF of equivalent fatigue loading
[-]
In contrast to the preceding aspect the dynamic amplification of the wave loading is a
remarkable issue. Figure 9.16 presents the dynamic amplification of the damage
equivalent overturning moment at the mudline for pure wave loading according to
lumped load cases similar to those from Table 8.3. When the turbine is parked, no
aerodynamic damping is present and the dynamic amplification factor (DAF) grows
8
7
6
5
turbine parked:
0.5% structural damping
Hs = 4.5 m
Tz = 6.4 s
p = 27 h/year
3
2
1
0
0.10
Hs = 0.5 m
Tz = 3.4 s
p = 1300 h/year
Hs = 1.7 m
Tz = 4.4 s
p = 440 h/year
4
production: structural and aerodynamic damping
0.12
0.14
0.16
0.18
0.20
0.22
spectral peak frequency [Hz]
Figure 9.16: Dynamic amplification of the damage equivalent bending moment at
mudline of the Opti-OWECS monopile due to pure sea state loading
207
proportionally to the peak frequency of the Pierson-Moskowitz spectrum. Excessive DAF
values up to 7 are observed since the spectral peak frequency is close to the
fundamental natural frequency of 0.29 Hz.
During production when aerodynamic damping is active the DAF varies only between
2.15 and 3. Obviously two opposing effects are present. The greater aerodynamic
damping ratio occurring at lower wind speeds (Figure 7.11) compensates the larger
dynamic amplification for higher peak frequencies as seen without aerodynamic
damping. For illustration the parameters of three sea states are given. The high dynamic
amplification at parking is associated with small but very frequent waves whilst the sea
states during the few days with severe storms exhibit a lower DAF.
Stress range distributions and effect of availability on fatigue response
Figures 9.17 and 9.18 plot the cumulative Rainflow counts of the bending stresses at
two locations, close to the tower top and below the mudline. In addition, the damage
equivalent stress ranges for a S-N curve with inverse slope of 4 are shown.
Five different combination of omnidirectional and collinear loading are investigated:
No. 1: Response on pure wind excitation during production at 100% availability
No. 2: Pure wave response without aerodynamic damping
No. 3: Wind and wave response at 100% availability, i.e. combined wind and wave
response in the production wind speed range and pure wave response below cutin and above cut-out wind speed
No. 4: Wind and wave response at 85% availability, i.e. as No. 3 but pure wave
response without aerodynamic damping when turbine unavailable
No. 5: Wave response in the production wind speed range with aerodynamic damping
generated by operation of the rotor in a steady wind field
No. 3 and No. 4 represent actual fatigue conditions with different availability whilst the
other cases are fictitious. A comparison of the cases 1 and 5 reveals the relative
contribution of the aerodynamic and the hydrodynamic loading. At the monopile the
wave loads dominate while the opposite holds for the upper tower part. It should be
noted that the wave loading of the tower is purely caused by the inertia loads of the
dynamic response.
The difference of conditions 2 and 5 is an indication for the dynamic amplification of the
wave response in the production wind speed range if no aerodynamic damping is
present. Both cases are not fully compatible since case 2 includes also the wave
response below cut-in and above cut-out winds. The ratio of total number of cycles and
total duration corresponds to 95% of the fundamental eigenfrequency of the monopile
structure. This is again a consequence of the large dynamic response.
The total number of cycles of the cases with wind contribution, i.e. No. 1, 3, and 4, is
driven by the excitation with the rotor frequency 1P and its multiples. The figures
correspond to approximately 4P and 2P at the upper tower part and the monopile,
respectively.
As mentioned in Section 8.1 both an upper and a lower availability should be compared.
For the present case the contribution of the wave response without aerodynamic
damping is so significant that for both cross sections a lower availability (No.4) results
in a shorter lifetime than full availability (No.3).
208
Chapter 9
60
1
stress range [MPa]
50
4
damage equiv. stress range [MPa]
60.00
3
2
40
30
1
5
2
20
10
tower 10.6 m below tower top
50.00
40.00
30.00
20.00
10.00
design lifetime 20 years
1.E+06
number of cycles
1.E+07
1.E+08
log(N) [-]
No.5
1.E+05
No.4
1.E+04
No.3
1.E+03
No.2
0.00
1.E+09
1.E+09
No.1
0
1.E+02
No. 1: pure wind (100% availability)
No. 4: wind & waves (85% avail.)
No. 2: waves & no aerodyn. damping
No. 5: waves & aerodyn. damping (100% avail.)
Figure 9.17: Cumulative Rainflow counting and damage equivalent
stress ranges of bending stresses at a location near the tower top
of the Opti-OWECS monopile design
80
80
damage equiv. stress range [MPa]
2
60
50
4
40
3
30
5
20
1
monopile 5.5. m below mudline
60
50
40
30
20
10
design lifetime 20 years
1.E+05
1.E+06
number of cycles
1.E+07
log(N) [-]
1.E+08
1.E+09
1.E+09
No. 1: pure wind (100% availability)
No. 2: waves & no aerodyn. damping
No. 5: waves & aerodyn. damping
Figure 9.18: Cumulative Rainflow counting and damage equivalent
stress range of bending stresses at a location below the mudline
of the Opti-OWECS monopile design
No.5
1.E+04
No.4
0
1.E+03
No.3
0
1.E+02
No.2
10
70
No.1
stress range [MPa]
70
209
Lateral response and misalignment of wind and waves
So far we have shown only results for the fore-aft response and for collinear wind and
wave loading. Lateral wind speed variations, variation of the aerodynamic loading of the
different rotor blades, mass imbalance of the rotor and especially the negilible
aerodynamic damping of the lateral motion cause a significant aerodynamic response
in the lateral direction. Specific to offshore wind energy converters is the lateral
excitation by waves that are misaligned to the rotor orientation and the spreading of
wave energy with respect to the mean wave direction (short-crestedness).
Figure 9.19 presents the distribution of the damage equivalent stress range (µ = 4) along
the circumference of the critical cross section of the monopile below the mudline. In all
considered cases the maximum response occurs in the fore-aft direction. At the pile the
lateral aerodynamic response, No. 1, accounts to approximately one third of the fore-aft
response. In the upper tower part (not shown) this ratio increases approximately to one
half. If long-crested waves, No. 2, approach always from the same direction, no lateral
response occurs. Therefore the lateral response of the case No. 4 with collinear wind
and wave loading at 85% availability is equal to the lateral response for pure wind
loading. Most realistic is the case No. 4a where a misalignment of the mean wind and
wave direction according to the distribution from Figure 4.18 is taken into account. The
lateral response is more than double compared to the collinear condition but the fore-aft
response is not altered significantly.
As explanation of this phenomenon the distribution of the damage equivalent stress
range is shown for five different angles between the wind and the wave direction (Figure
9.20).6 Up to a misalignment of ± 30o, which represents 82% of the time, the total
response in the wave direction is not altered significantly but the distribution along the
circumference flattens. For large misalignments of 60o and 90o the absence of the
aerodynamic damping increases the wave response and it exceeds the maximum total
loading for collinear conditions.
Short-crestedness has not been investigated, yet. The above results indicate however
similar behaviour for typical conditions where the wave energy is distributed relatively
narrow around mean direction (Section 4.1.2).
Directionality
Obviously, the magnitude and frequency of the wind and wave loading at a particular site
depends on the direction and the associated fetch length and water depth. Figures 4.16
and 4.17 show the variation of the significant wave heights for different directions and
prevailing wind and wave directions close to the NL-1 site. Consideration of directionality
in time domain simulations is cumbersome and care is required to achieve conservative
results. Figure 9.21 compares the damage distribution along the circumference of the
pile for four cases which are discussed in order of increasing sophistication.7 Firstly,
omnidirectional loading is applied. Independently from their actual direction wind and
waves are assumed collinear and they approach always from the prevailing wind
direction. This method is clearly conservative with respect to the maximum damage and
6
Due to practical considerations of the computation here the wave direction is maintained and the wind
direction and turbine yaw angle are varied.
7
In contrast to the two previous figures we are dealing with a damage distribution because this enables
the convenient re-distribution of the fatigue results from the different load sectors.
210
Chapter 9
o
o
wave direction 0
wind direction, yaw angle 0
(No.1 - No. 4)
70
No. 1: pure wind (100% avail.)
60
50
40
30
20
10
0
No. 2: waves & no aerodyn.
damping
No. 4: collinear wind & waves
(85% avail.)
No. 4a: misaligned wind &
waves (85% avail.)
( wind direction classes: -90o,
o
o
o
o
o
o
o
-60 , -30 , -15 , 0 , 15 , 30 , 60 )
monopile 5.5.m below mudline
Figure 9.19: Distribution of damage equivalent stress range on the pile
circumference for different loadings (µ = 4)
(See Figure 4.18 for misalignment of wind and waves for loading No. 4a.)
wave direction 0o
70
60
misalignment of wind direction
& turbine yaw angle
with respect to wave direction
50
collinear
40
- 15 deg yaw
30
20
- 30 deg yaw
10
- 60 deg yaw
0
- 90 deg yaw
monopile 5.5.m below mudline
85% availability
Figure 9.20: Distribution of damage equivalent stress range on the pile
circumference for different misalignment of wind and wave direction (µ = 4)
211
N
1.50
omnidirectional loading
NW
Damage [-]
1.25
NE
1.00
omnidirectional loading
distributed by wind rose
0.75
0.50
directional loading
(collinear wind & waves)
0.25
W
E
0.00
SW
SE
directional loading (noncollinear wind & waves)
85% avaialibility
(design lifetime 20 years)
S
Figure 9.21: Distribution of damage on circumference of monopile for four
different models for the directionality of the wind and wave loading
should be considered as the default analysis. Secondly, the damage from the previous
case is distributed over the perimeter according to the distribution of the mean wind
direction. This approach results in the most regular damage distribution but also in the
minimum damage in the prevailing direction. Next, each 21 lumped load cases are
generated for 6 wind direction sectors of 30o ± 180o (Table B.2 in Appendix B) and 224
cases with collinear loading are simulated for 40 minutes each.8 Figure 9.19
demonstrated an increase of the lateral response when wind and wave direction are
considerably misaligned. Therefore in the final approach the maximum damage is
calculated from the lumped load cases for each of the six load sectors and subsequently
the relative distribution of the damage on the circumference corresponding to case No.
4a from Figure 9.19 is accounted for. This procedure yields the most realistic damage
distribution as long as short-crestedness is not considered additionally. The maximum
damage is only half in comparison to the omnidirectional loading but the required quality
of the correlated wind and wave data and computational effort are high.
9.6.2 Extreme response
It is beyond the scope of the thesis to discuss the extreme response of offshore wind
energy converters in general and for all reference designs. Instead examples are given
for the extreme loading during power production and during 50 years survival conditions.
Extreme loading during power production
The establishment of the extreme aerodynamic and hydrodynamic conditions during
power production has to consider in principle the dynamic response rather than the
8
The directional wind distribution is considered because it is more pronounced than the wave distribution
in Figure 4.17. The fatigue of the steel pile does not depend on the sign of the mean loading hence
opposite sectors are treated together.
212
Chapter 9
extreme environmental parameters. So far, existing standards, as for instance the
Germanischer Lloyd guidelines, propose the combination of an extreme aerodynamic
event with a moderate hydrodynamic condition and vice versa. This is appropriate only
if one of the two effects is dominating which is indeed the case at the following example
of the Opti-OWECS monopile design. If wind and wave induced response are of similar
magnitude, various circumstances are imaginable which may lead to a higher response
than the cases with one dominating load component. Further development of structural
reliability methods is required for safe design of multi-megawatt OWEC at demanding
sites.
Germanischer Lloyd proposes the offshore load case E1.1 as ‘Extreme operating gust
of 13 m/s with directional change and a reduced wave loading 1.32 · Hs at a sea state
associated with the mean wind speed’.
The characteristic bending stresses at the pile section below the mudline for different
loadings are shown in Figure 9.22. The stress levels are far lower than the yield and
buckling strength, so, fatigue and the dynamic requirements are the design drivers for
the support structure (Figures 10.11 and 10.13). The total response is driven by the
aerodynamic contribution owing to the five times greater magnitude of the wind turbine
induced response compared to the wave case. Moreover, above the rated wind speed
the responses show a reverse dependency on the mean wind speed.
For this pitch regulated turbine the highest wind response occurs at a mean wind speed
at or below the rated wind speed. The controller is not fast enough to increase the pitch
angle according to the instantaneous wind speed and a high overshoot of the tower top
motion takes place. Furthermore the magnitude of the directional change, as specified
by Germanischer Lloyd, is very large in this range (Table 4.2).
The negative overshoot of the tower motion at low mean wind has several reasons.
During the first few seconds of the gust the axial thrust loading increases but is followed
then by a rapid decease below zero due to the large yaw misalignment of the flow and
the initiated shut-down of the turbine. The flow conditions during pitch action and the
employment of the mechanical brake can cause a transition from the turbine to the
propeller stage (Figure 9.23). After comparison of different settings, the combination of
a moderate pitch rate of 5o/s and braking with twice the nominal torque was chosen.
The wave loading increases almost linearly with the mean wind speed. This could be
expected for the quasi-static loading since it is approximately proportional to the wave
height. The dynamic wave loading behaves similarly owing to the almost constant DAF
shown in Figure 9.16.
The operational load case E1.1 produces the highest extreme loading for the design. In
comparison the maximum characteristic stresses associated with the extreme gust load
case E2.1 and the extreme wave load case E2.2 at 50 years survival conditions are only
in the order of 52 and 60 MPa, respectively.
Extreme loading during 50 years storm conditions
In comparison to the other reference designs the Utgrunden case is most interesting
with respect to the extreme wave loading in shallow waters and the ice loading. The
design wave height of 6.4 m corresponds to the breaking wave height for a water depth
of approximately 8.25 m which is within the range of water depths at the site. Actually
the wind farm is erected on a subsea ridge and wave breaking may occur. Also the
monopile with its soft-soft characteristics is sensitive to dynamic wave loading.
213
characteristic stress [MPa]
100
total loading (min,
max)
50
aerodynamic
loading
(min, mean, max)
hydrodynamic
loading
(min, max)
0
-50
-100
5.0
7.5
10.0
12.5
15.0
17.5
20.0
22.5
25.0
Wind speed at hub
25
20
15
12
Pitch angle −− and wind direction −
90
45
0
−45
−90
Axial tower top force
500
250
0
−250
−500
−750
Rotor speed
[rpm]
22
11
0
Aerodynamic power
4
[MW]
[kN]
[deg]
[m/s]
Figure 9.22 Extreme stresses of the Opti-OWECS monopile below mudline for the
load case E1.1: ‘Extreme operating gust and reduced wave’ at different wind speed
2
0
−2
0
2
4
6
8
10
time [s]
12
14
16
18
Figure 9.23: Time series for load case E1.1: Extreme operating gust at a
mean wind speed of 12 m/s (no hydrodynamic loading)
20
214
Chapter 9
E 1.1 Operating gust
E 2.2 Design wave
E 2.2 Design wave pseudo dynamic
E 2.2 Sea state
E 2.2 Sea state pseudo non-linear
E 2.4 Static sea ice
0.00
5.00
10.00
15.00
20.00
25.00
30.00
Design load [arbitrary units]
Aerodynamic
Hydrodynamic
35.00
40.00
45.00
Sea ice
Figure 9.24: Comparison of load magnitudes at the monopile of the extreme load
cases at the Utgrunden wind farm in 10 m water depth (MSL)
As a matter of fact both conditions are not fully compatible with common engineering
analysis methods. On one hand the calculation of the dynamic wave loading should
consider the randomness of the extreme sea state which is generally only possible by
the use of the linear wave theory in its pure form or corrected by an empirical
modification, e.g. the Wheeler stretching (Section 4.1.2).
On the other hand non-linear wave theories are commonly capable only of a single
deterministic wave with a prescribed height and period rather than an entire sea state.
Of course, it is possible to calculate the dynamic response on a non-linear wave in the
time domain. However, the response is very sensitive to the chosen wave period and
it is unclear whether the transient or the steady-state response is more realistic.
Calculations for the present monopile design showed only a moderate dynamic
amplification of the transient response in the order of 1.5 during the first wave cycles of
long waves but unrealistic high dynamic amplification factors of up to 16 (!) for the
steady state response of shorter waves when the second harmonics coincides with the
fundamental natural frequency.
During the preliminary design phase of the Utgrunden project a pragmatic and
conservative way was followed to estimate the extreme wave loading [9.15].
A generalized dynamic amplification factor (DAF) of the extreme wave loading of 3.4
was determined from the comparison of simulations with and without structural
dynamics. The DAF was related to the median value of the extreme overturning moment
of 10 simulations of the extreme sea state with 3 hour duration each. In parallel, a nonlinearity operator of 2.1 was derived from the comparison of the quasi-static overturning
moment according to a 10th order Stream function wave (Figure 5.15) and from
application of linear wave theory with Wheeler stretching. By this two new load cases
were synthesised. The so-called ‘E2.2 Sea state - pseudo non-linear’ case comprises
the nearly linear response from the sea state loading multiplied by the non-linearity
operator. Secondly, load case ‘E2.2 Design wave - pseudo dynamic’ combines the
quasi-static, non-linear wave loading multiplied by the dynamic amplification factor from
the extreme sea state analysis.
215
Figure 9.24 compares the relative magnitudes of the design loads of the different load
cases and load origins. The static sea ice loading for 40 cm of ice with a crushing
strength of 2 MPa dominates the monopile bending below the mudline. In contrast to the
Opti-OWECS monopile the loading at extreme wind speed conditions is more severe
than during operation. The magnitude of the extreme sea state response is similar to
that of the design wave. Multiplication of the design wave loading with the large value
of the above-mentioned dynamic amplification factor results in the most onerous wave
loading of the two constructed load cases.
The above-mentioned treatment is conservative since distinct loading effects and
frequency ranges are associated with the non-linearity and the dynamic amplification of
the response.
The non-linear contribution of the response is mainly related to the quasi-static loading
by the non-linear drag at the low frequency, high energy part of the wave spectrum. In
contrast, the dynamic response mainly occurs in the vicinity of the fundamental
eigenfrequency located in the high frequency, low energy tail of the wave spectrum. In
this frequency range the wave loading is inertia dominated and almost linear. Figure 8.6
shows typical wave energy and response spectra for such a case.
Instead of the multiplication of the non-linear, quasi-static response with the dynamic
amplification factor it would be more appropriate and less conservative to add the
additional dynamic response to the quasi-static, non-linear response. Reference [9.16]
discusses this problem in more detail and compares two suitable engineering methods.
9.7
Comparison of design drivers for the support structure
At the end of the chapter an overview of the design drivers and some complementary
results for the support structures of the seven reference designs is given. Further
information can be found in previously published material [9.1, 9.5, 9.15, 9.17 - 9.19].
Soft-soft and soft-stiff monopiles at the wind farm Lely in the IJsselmeer
The Lely monopiles A2 and A3 were designed with respect to the dynamic requirements.
The highest extreme loading with a utilisation of approximately 50% is related to static
ice loading at this fresh water site in combination with an extreme wind speed. Only wind
turbine induced fatigue was checked in the original design calculations [9.3].
Dynamic analysis of the combined wind and wave response at the soft-stiff design A3
showed two times higher equivalent fatigue stresses than for pure wind loading. The
fatigue safety factor is however still large. The monopile A2 with an indented soft-soft
characteristics is subject to an extensive wave induced fatigue if the soft-soft behaviour
would actually exist. The total aerodynamic and hydrodynamic fatigue loading is about
four times higher than for the stiffer companion A3. This amazing result is caused by the
low eigenfrequency, very high hydrodynamic participation factors due to a too small
penetration depth (Table 9.4) and the short waves on the IJsselmeer.
The fatigue performance of the fictitious design A2-mod is significantly improved; the
equivalent loading is only half of design A2.
The foundation sensitivity of design A2 is unacceptable while A3 and A2-mod show
moderate sensitivity. The reasons for the actually much stiffer behaviour of the A2
design are still unclear and are subject to further investigations.
216
Chapter 9
Soft-soft monopile at the Utgrunden reef
The weight per unit length of the monopiles at Utgrunden and at Lely A3 is equal,
nonetheless, a wind turbine with three times higher rating and an approximately 50%
greater hub height (MSL) is employed 6 years later. This was possible since the
monopile was better balanced with respect to the dynamic and strength requirements
and was designed as one unit together with the tower. The variable-speed, variable-pitch
regulated, three-bladed turbine shows lower loads in relative terms than the active-stall
regulated, two-bladed Lely design.
The Utgrunden support structure is governed by the extreme sea ice and wind speed
condition. The aerodynamic contribution dominates the fatigue loading because the
wave regime is moderate and the foundation is relatively stiff and shows very low
hydrodynamic participation factors (Table 9.4). The design is appreciably insensitive to
variation in the soil properties.
Gravity lattice tower at NL-5 site with WTS 80 turbine
Experience from the offshore technology was used for the design of the lattice tower at
the most demanding site. The transparent structure attracts only low wave loads and
little dynamic amplification is observed, thanks to the stiff characteristics. The dynamics
are driven by the high amplification of the aerodynamic loads at the strong 2P excitation
of the blade passing and the stress concentration at the tubular joints. A fatigue analysis
showed the need for major structural reinforcements in the upper truss.
At extreme conditions the aerodynamic base shear account for only 13% of the wave
and current loading. The size of the caissons at the three legs is determined by the
sliding resistance and the overturning moment. The foundation sensitivity is moderate
for the first fore-aft mode but quite high for the second mode.
Soft-stiff gravity monotower at the Baltic Blekinge site with the WTS 80 turbine
The gravity monotower is, likewise to the lattice tower, dominated by the dynamic
response at the blade passing frequency of the turbine. The damage equivalent
overturning moment for combined wind and wave loading is only 4% higher than for pure
wind loading. The gravity base experiences a very considerable heave force during the
passage of the extreme design wave which can be countered only by 3,000 t of ballast.
Like the lattice tower, the foundation sensitivity is moderate for the first fore-aft mode but
quite high for the second mode.
Soft-soft monopile at NL-1 site with the WTS 80M turbine
The final design solution of the Opti-OWECS project was found after an integrated
design process (Chapter 10) and is dominated by wind and wave induced fatigue. The
soft-soft characteristics is a compromise between the wind turbine and support structure
requirements and is possible only with a higher rotor speed of the turbine. A tapered
monopile with a small water-piercing cross section is needed to keep the wave loading
within acceptable bounds. The original design from the Opti-OWECS project was found
to require some small reinforcement after considering a poor availability of 85% of the
turbine and lower structural damping of only 0.5%. The increased loading is purely
caused by a higher contribution of dynamic wave response at parking. A water depth of
20 m (MSL) at good soil conditions clearly limits the viability of the concept.
The highest extreme loads occur during turbine operation at partial load conditions when
wave loading is low. The foundation sensitivity is low for medium dense to dense sand.
Design Optimisation by Means of Tailored Dynamics
217
CHAPTER 10
DESIGN OPTIMISATION BY MEANS OF TAILORED DYNAMICS
Tailoring the dynamics of an offshore wind energy converter can offer an effective
design optimisation. The present chapter demonstrates this statement by following the
design of a large-scale offshore wind farm (Figure 10.1) through three successive stages
of the design process: feasibility study, conceptual design and detailed design (Sections
10.1 to 10.3).
The chapter integrates results of previous chapters. Consideration is given to the
integrated design approach from Chapter 3, modelling appropriate to the stages of the
design process (Chapters 4 to 6), the analysis methods for aerodynamic and
hydrodynamic fatigue introduced in Chapters 8 and the interrelation of dynamics and
design as discussed in Chapter 9.
In addition to the top-down or vertical approach following the design through its
development process, also a thematic or horizontal structure is given to the chapter. In
each design phase we will present and demonstrate a suitable treatment of the
essentials of dynamics and design:
• environmental description
• dynamic characteristics
• aerodynamic damping
• foundation uncertainty
• fatigue and extreme response
• economic evaluation
Figure 10.1: Artist impression of the Opti-OWECS design solution
Chapter 10
global data for entire regions:
Weibull wind speed
distribution, empirical closedform relations for wind and
waves
separate models for wind
energy converter in calm
water and support structure
with top mass (simplified
design tool)
Rayleigh quotient or similar
methods, concept of apparent
fixity length for piled
foundations
closed-from estimate,
average value during power
production
empirical uncertainty range
for fundamental
eigenfrequency
Structural model
Dynamic
characteristics
Aerodynamic
damping
Foundation
uncertainty
Feasibility study
Finite Element analysis with
properties of soil layers from soil
investigations, systematic variation
of all structural and environmental
uncertainties
consideration by integrated, nonlinear time-domain simulations
variation of estimated values of soil
parameters from site investigation
determination of
aerodynamic damping for
different wind speeds
empirical variation of soil
stiffness
integrated OWEC model incl. an
aero-elastic wind turbine model
and a Finite Element model of the
support structure
high quality met-ocean database for
the specific site: scatter diagram for
wind speed and wave parameters
Detailed design
Finite Element analysis with
indicative soil properties
separate modelling for wind
energy converter in calm
water and support structure
with top mass (standard
design tools)
specific site data: Weibull
wind speed distribution and
wave scatter diagram
Conceptual design
Table 10.1: Modelling and analyses of OWEC dynamics and design within the integrated design approach
Environmental
description
218
219
long-term superposition of
separate wind and wave
fatigue analyses in time
domain and frequency
domain, respectively
spreadsheet like analyses or
scaling from similar condition
by rule of thumb
global cost model, screening
of possible sites with a
Graphical Information System
Fatigue
Extreme response
Economics
Feasibility study
detailed engineering cost
model
standard design tools: quasistatic analysis of wave
response plus Dynamic
Amplification Factor,
dynamic analysis of wind
response
short-term superposition of
separate wind and wave
fatigue analyses in time
domain and frequency
domain, respectively
Conceptual design
direct evaluation of component costs
OWEC design tool: structural reliability
analysis (if design driver,
else quasi-static or dynamic analysis)
non-linear time domain simulation
of simultaneous wind and wave
response for lumped load cases and
Rainflow counting
(if design driver, else simplified
method)
Detailed design
Table 10.1 (cont.): Modelling and analyses of OWEC dynamics and design within the integrated design approach
220
Chapter 10
Table 10.1 provides an overview on the proposed approaches. While the first five
aspects have also been addressed in the previous chapter, the economic evaluation is
introduced here for the first time. For completeness also the structural modelling is
mentioned in the table.
The design solution is very similar to that of the Opti-OWECS project, an offshore wind
farm with 100 units of 3 MW with the WTS 80M wind turbine and a soft-soft monopile at
the Dutch North Sea site NL-1 [10.1]. In the scope of this thesis the methods used and
some assumptions are updated and the design process is repeated. Therefore slightly
different results are established.
10.1Feasibility study
The purpose of the feasibility study is a rough estimate of investment and energy costs
for a number of potential sites and associated wind farm designs. In our case one site
and two preliminary design concepts are the outcome which are further developed
during the conceptual design phase.
10.1.1
Treatment of dynamics and design
Environmental description
During the early design stage correlated wind and wave data or even any specific data
for the particular site might not be available and one has to use one of the following
simplified environmental models listed in order of increasing complexity.
• Weibull distribution of the mean wind speed,
Neumann-Pierson relation, (10.1), giving significant wave height, Hs ,and zerocrossing period, Tz, for fully developed sea states in deep water as function of the
hourly mean wind speed at 10 m height, V1h, (Figure 4.16) [10.2, 10.3]. One should
be aware of the conservatism of the formulas when they are applied to fetch limited
sites in shallow waters.
0.243 2
5.48
Hs '
Tz '
V1h
V1h
g
g
(10.1)
where:
g gravity acceleration
Weibull distribution of significant wave height in European waters derived from
voluntary ship observers [10.2] or similar data from the Baltic Sea Handbook [10.4]
in combination with an empirical relation between significant wave height and zeroup crossing period (wave steepness)
two-dimensional Hs - Tz scatter diagram
In the last two cases a monotonic wave height - wind speed relation can be derived by
equating parameters with the same cumulative probability of occurrence.
Indicative values for extreme environmental conditions can be gained, e.g. from the
Guidelines of the UK Department of Energy [10.5] or the Baltic Sea Handbook [10.4].
Both for fatigue and extreme conditions the shallow water breaking wave limit (Section
4.1.2) should be accounted for in order to avoid unrealistic large wave heights and too
221
long wave periods. In all cases wind and waves are considered as collinear and
omnidirectional.
Dynamic characteristics
Various textbooks provide closed-form estimates for the fundamental eigenfrequency
of the support structure. Approximative numerical methods like the Rayleigh quotient
can be implemented easily within a spreadsheet that can also be used for a simplified
strength analysis.
Consideration of the soil characteristics can be more important than the detailed mass
and stiffness distribution along the structure. For gravity systems the foundation stiffness
can be estimated by Eqn. (5.6) whilst for piled foundations the concept of the apparent
fixity length (Section 5.2.2 and 9.3.1) is extremely useful.
The hub height is a design parameter with strong influence on the dynamic
characteristics. At offshore sites a lower height than on land is advantageous. During
the Opti-OWECS project the minimum hub height was determined as:
+
+
+
+
+
+
tidal amplitude
surge
extreme crest height (approximately b of extreme wave height)
1.5 m air gap according to Germanischer Lloyd
platform height (0.5 m)
4.5 m minimum blade clearance
rotor radius
minimum hub height above MSL
(10.2)
The minimum blade clearance was established after consultation with Germanischer
Lloyd and in reference to a local German building rule [10.6]
Aerodynamic damping
As discussed in Section 8.2 the assumption of only two values of the aerodynamic
damping ratio, corresponding to the idle condition and the average behaviour during
power production of the OWEC, is reasonable at this stage of the design process. The
closed-form relation (5.12) can be evaluated. However, the dynamics of the pitch control
system (if any) and the effect of the aerodynamic drag on the damping of stall regulated
turbines is not included in this way.
During the preliminary design only rather general soil condition data are available.
Instead of detailed parameter studies on the foundation behaviour it is more meaningful
to assume directly an uncertainty range for the fundamental eigenfrequencies based
upon experience. Without further knowledge of the actual conditions a tolerance of 15%
relative to the design eigenfrequency is recommended.1 Such a large uncertainty margin
is actually required with respect to the rough assumption and limited data of the early
design stage. This conservatism could, however, hinder to pinpoint the really interesting
design concepts.
1
Positive and negative eigenfrequency tolerance are stated with equal magnitude since the
eigenfrequency depends decreasingly on the foundation stiffness. For instance, a variation from 50% to
300% of the reference soil stiffness may result in approximately ±5% change in the eigenfrequency
(Figure 9.13).
222
Chapter 10
±.1P
0-
0- 0+
design
range fD
0+
f
1P soft-stiff
design
soft-soft
±.NbP
Nb P
range fD
stiff-stiff
design
range fD
.1P .NbP exclusion range at excitation frequency 1P and NbP
0 - 0 + uncertainty range
Figure 10.2: Design ranges for the fundamental eigenfrequency
fD
fR,min
fD
fR,max
fD
fR,min
fD
fR,max
#
$
#
1
soft&soft design
( 1 % ζ 1P ) ( 1 % η& )
1
( 1 & ζ1P ) ( 1 & η% )
Nb
soft&stiff design
( 1 % ζ N P ) ( 1 % η& )
b
$
Nb
%
( 1 & ζN P ) ( 1 & η )
stiff&stiff design
(10.3)
b
design value of eigenfrequency
where: f D
fR,max , fR,min upper, lower rotor frequency
Nb
ζ1P , ζN
η& , η%
number of blades
bP
exclusion range around fR and Nb fR
structural uncertainty of eigenfrequency
We denote these dimensionless, positive tolerances as η- and η+. Even under bad
conditions the eigenfrequency should not fall into the exclusion ranges, ζ1P and ζNbP ,
around the rotor frequency, 1P, and blade passing frequency, Nb P. Figure 10.2
illustrates the three ranges for the design value fD of the fundamental eigenfrequency as
well as the exclusion and uncertainty ranges.
Once the first eigenfrequency calculations for a preliminary design are available, the
tolerance could be related to the structural behaviour. The difference of the
eigenfrequencies between a flexible and a rigid foundation gives a good indication of the
223
sensitivity of the design. For a specific site, a foundation tolerance η+ of approximately
half of this distance should be a safe estimate, substituting the previously mentioned
general value of 15%. For individual tolerances the design ranges of the fundamental
eigenfrequency, fD , can be determined by Eqn. (10.3). Default design ranges for two
and three-bladed rotor concepts can be read from Table 10.2.
Table 10.2: Indicative design ranges for the first eigenfrequency, fD , relative
to the rotor frequency, fR ( ζ1P = 5%, ζNbP = 10%, η- = η+ = 15% )
Rotor
soft-soft
soft-stiff
twobladed
fD
fD
threebladed
fR,min
# 0.83
1.24 #
fD
fR,min
fR,max
fD
fR,min
stiff-stiff
# 1.58
2.61 #
# 2.37
3.92 #
fD
fR,max
fD
fR,max
The ratio between wind, wave and ice loads should be studied for a couple of extreme
load cases.2 Concerning the fatigue loading the simplified analysis based upon the
superposition of equivalent long-term fatigue loads associated with the wind and wave
response is most suitable (Section 8.2).
In such an early design stage the required design data for time domain simulations of
the aerodynamic response might not be available or the level of detail involved in such
computations might not be appropriate. In such cases it could be worthwhile to derive
the loads by a rule-of-thumb from previous designs or different site conditions. Distinct
support structure types and different hub heights can have a large impact on the
fundamental eigenfrequency and consequently on both extreme and fatigue loading.
This should be considered when estimating the preliminary load assumptions.
Economic evaluation
For a commercial development, economic criteria related to the entire offshore wind
farm must be qualitatively evaluated right from the beginning. Various established
approaches exist in the financing field for the determination of the profitability of an
investment. Classical methods for Discounted Cash Flow Analyses include the Net
Present Value Method 3 and the Yield Method 4 [10.7].
2
For instance, referring to Table 4.4 the load cases E1.1, E2.1, E2.2 and if relevant E2.4 according to
Germanischer Lloyd could be analysed.
3
The expected annual income from an asset is discounted, i.e. reduced, to allow for the delay in
receiving that income, using a prescribed rate of interest. If the total of these discounted annual amounts
exceeds the capital sum needed to buy the asset now, the investment may be considered profitable.
4
One determines at what rate of interest the total discounted annual amounts of income would be exactly
equal to the capital sum needed to buy the asset now. If the rate of interest, or yield, is greater than the
market rate, the investment may be regarded as profitable.
224
Chapter 10
In this thesis the economic question is stated with a slightly different motivation.
Economics, or more precisely investment and energy costs, are considered as important
criteria in the evaluation of the performance of a particular design. Moreover, these
economic quantities are used for the choice between different design options and the
(global) comparison between the cost of offshore wind energy and other energy sources.
For such engineering applications the method of the levelised production cost has been
established by the International Energy Agency (IEA) [10.8].
The levelised production cost (LPC) is defined as the average cost per produced energy
unit over the economic lifetime of the plant. The total net energy output and the total cost
over the economic lifetime, i.e. capital, running and decommissioning costs, are
discounted by a chosen test discount rate to the start of the operation. Then the LPC is
obtained as ratio of discounted annual cost and net annual energy output.
This approach is based upon a constant amount of capital, the annuity, that has to be
returned to the lender each year. Energy yield and all other costs except the initial
investment can vary in principle during the lifetime; however, their effect on the energy
cost is levelised over the lifetime. Assuming constant net annual energy yield and
constant annual operation and maintenance cost results in Eqn. (10.4) and (10.5).
LPC '
where:
I
TOM
%
a Ey
Ey
LPC
levelised production cost
I
investment incl. interest during construction
a
annuity factor a '
1
1&
(1 % r )
1
r
r
test discount rate
ne
economic lifetime in years
TOM
total levelised annual ' downtime cost '
Ey
annual energy yield
TOM ' OM %
(10.4)
DC
a (1 % r)
where:
ne
ne
(10.5)
OM annual operation & maintenance cost
DC net decommissioning cost
Typical values for the economic parameters used at commercial wind farms can be
found in Table 10.3 [10.9]. In accordance with the first large-scale project in Denmark
we apply a test discount rate of 5% and 20 years loan.
225
Table 10.3: Typical test discount rates and repayment periods
Test discount
rate †
Repayment
period
Annuity
factor a
Denmark
7%
20 years
10.6
Germany
varies, 5 %
upwards
10 years
7.7 or lower
The Netherlands
4 to 5 %
10 years
7.7 to 8.1
United Kingdom
developer’s
choice
15 years
5%
20 years
IEA Recommendation ‡
†
‡
12.5
rate at which the nominal rate exceeds the inflation rate
for comparison of different energy sources
Cost models
The LPC can be evaluated even by hand if the different quantities are known. An
estimate of the investment cost is however a difficult task which can be eased by
general cost models.
During this research three of such models were used. In 1995/96 Köhler and Pauling
developed the SCOptiM (Simple Cost Optimsation Model) code [10.10 - 10.12]. During
Opti-OWECS project a more advanced cost model was applied [10.13].
For this thesis the simplified design tool MonOpt (Monopile Optimsation) was
developed. Features of the MonOpt code include:
•
Finite Element like model of cantilever structures
calculation of arbitrary number of eigenfrequencies
•
•
wave force calculation with linear wave theory and Morison equation
•
spectral analysis of hydrodynamic fatigue
direct quadratic superposition of long-term wind and wave fatigue
•
•
extreme wind loading of the tower
•
static ice loads according to [10.14]
linear strength and buckling analysis according to DIN 18800
•
•
support structure cost estimate according to Opti-OWECS cost model
•
energy cost evaluation according to IEC recommendations.
Implementation has taken place in MATLAB™ 5.3.
10.1.2
Preliminary design of the Opti-OWECS wind farm
Design objectives
As illustration of the optimisation approach the development of a 300 MW offshore wind
farm is taken from the Opti-OWECS study [10.1]. It would be beyond the scope to repeat
here a treatment of similar extension as in the European project.
226
Chapter 10
We focus on the dynamics of the support structure and the wind turbine and take the
solutions for the grid connection and operation and maintenance from the study. Also,
the same design concepts for the support structure, installation and turbine are used.
Due to practical reasons we restrict us to the WTS 80 / WTS 80M machine, rated 3 MW
with a two-bladed, 80m rotor and variable-pitch, constant speed control.5 As starting
point the WTS 80 with a rated rotor speed of 19 rpm is considered. In the meantime,
larger turbines with a lower specific rating and lower dynamic loading due to variablespeed are under development and might be better suited. Such data was however not
available for this thesis.
The design goal is to achieve lowest levelised production cost based upon a test
discount rate of 5% and 20 years loan. All given energy cost are purely indicative, are
related to these financial parameters and neglect grid connection cost on land.
Pre-selected sites
Six distinctly different sites spread over northern Europe are selected as candidate sites.
Data of Table 10.4 are collected and harmonized from various sources given in the
project report [10.1]. The NL-1, NL-5 and SE-1 location are identical to those considered
in Chapter 9. The site UK-1 is located in the Wash bay on the British east coast at about
53o N. The German location DE-1 is near the city of Rostock and is chosen because of
its low water depth and close distance to shore. In fact, a nearshore site it would not be
suited for such a large offshore wind farm.
Compatibility of concepts and sites - Pre-selection of concept
Four generic support structure types are initially considered which are qualitatively
matched with the sites in Table 10.5.6
The monopile is suited good to excellent to all sites except of NL-5 where the large water
depth and demanding environment would require uneconomic dimensions. The possible
presence of boulders or rock covered by glacial sediment at the Swedish site could
complicate both driven and drilled installation. At present mainly a soft-stiff solution is
of interest. According to Table 10.2 a soft-soft monopile should have a design
eigenfrequency below 0.26 Hz if the WTS80 with a rotor speed of 19 rpm is used. High
aerodynamic and hydrodynamic fatigue makes such a design less attractive.
Different sites are compatible with the gravity monotower. At NL-5 the hydrodynamic
loads would require a heavy structure. During the Opti-OWECS project it was realised
that high waves in shallow waters generate very considerable heave forces which can
only be compensated by a large amount of ballast. Thus the suitability for DE-1 and DK1 is lower, as well.
Section 9.3.2 highlighted the high aerodynamic fatigue loads of designs with natural
frequencies in the vicinity of blade-passing frequency. In addition, the lattice tower is
subject to stress concentrations at the joints. Consequently, the lattice tower might only
have some chances at the most demanding NL-5 site. The Baltic sites neither fit to the
lattice tower nor to the tripod because the framework of smaller structural members is
too sensitive to ice loads.
The tripod or braced monotower corresponds ideally to the NL-5 site where a ‘stiff’
foundation with a ‘soft’ tower would be beneficial for the dynamics. Previously such a
5
See Section 9.1 for a further discuss of the particularities of this wind turbine concept.
6
Table 3.4-1 in [10.1] presents a comprehensive discussion of the pros and cons of various options.
227
design was analysed for such a location [10.2, 10.15]. At DK-1 the water depth is too low
for this concept.
During the Opti-OWECS project the gravity monotower showed significant economic
disadvantages compared to the monopile. We maintain this specific finding, knowing
that recent Danish research with own specific conditions indicates similar performance
[10.16]. The monopile is the best concept for all sites except NL-5 where the tripod is
more suited. Here we restrict ourselves to the possibilities offered by the monopile since
an optimisation of a tripod structure is beyond the scope of this chapter.
Evaluation of preliminary designs: Modelling
A parametric monopile design based upon the Opti-OWECS design solution together
with the WTS 80 machine with 19 rpm is analysed at the sites. The applied modelling
is briefly reviewed.
Wave scatter diagrams are not available for all sites. So the indicative Weibull
distributions of the significant wave height derived from voluntary fleet observations
[10.2] and data from the Blekinge feasibility study [10.17] are used. The Weibull
distributions for the significant wave height and the mean wind speed are correlated
according to the methodology described in Sections 8.2 and 10.1.1. A minimum
availability of 85% is assumed.
The structure was modelled Finite Element like with aids of the MonOpt code. A
constant fixity length of four pile diameters and a total penetration of 25 m are assumed.
The maximum bending moment was assumed to occur at a depth of two pile diameters.
The effective aerodynamic damping ratio during production is estimated at 2% of critical
damping for a first eigenfrequency of 0.3 Hz decreasing to 1% at 0.6 Hz.
Table 10.2 provides the ranges for the design fundamental eigenfrequency, fD:
soft-soft :
fD # 0.26 Hz
soft-stiff :
0.39 Hz #
fD # 0.50 Hz
Damage equivalent tower top fatigue loads are simulated in the time domain for four
values of the support structure stiffness and typical offshore turbulence intensities.
Interpolation gives the loads for other eigenfrequencies and other annual average wind
speeds. Five extreme load cases from Germanischer Lloyd are considered, for which
the wind loads are derived from simulations with a soft support structure. The MonOpt
code generates the hydrodynamic loads by spectral fatigue analysis or deterministic
analysis with linear wave theory and corrections for non-linear waves and dynamic
amplification.
Cost assumptions are based on the final Opti-OWECS design with a price level of winter
1997. Wind farm availability is corrected for both annual average wind speed and
distance from shore. Likewise electrical transmission losses are accounted for.
Evaluation of preliminary designs: Analysis
Comparison of the capital and energy cost in Figure 10.3 reveals the DK-1 and NL-1 site
as the favourites combining good wind conditions with moderate distance to shore and
environmental exposure. The Dutch site has slightly lower net energy yield (Figure 10.4)
and higher support structure cost owing to the greater water depth and higher waves.
Both disadvantages have to be paid by approximately 5% higher cost of energy.
The Baltic sites SE-1 and DE-1 range on intermediate positions whilst the British site
shows the highest cost due to three less favourable site parameters: low annual average
wind speed, considerable water depth and large distance to shore.
228
Chapter 10
Table 10.4: Site data for the feasibility study
Quantity
Annual average wind
speed at 60m
Water depth (MSL)
Distance from shore
Onshore distance to
HV grid point
Extreme mean wind
speed at 60 m
Extreme wave height
Design current
Surge
Tidal amplitude
Weibull scale factor
for Hs distribution
Weibull shape factor
for Hs distribution
Design ice thickness
Soil conditions
†
UK-1
The
Wash
8.2 m/s
NL-1
IJmuiden
9 m/s
NL-5
IJmuiden
9.5 m/s
DK-1
Horns
Rev
9.2 m/s
DE-1
Rostock
7.8 m/s
SE-1
Blekinge
8.4 m/s
20 m
30 km
17.5 m
15 km
5 km
25 m
50 km
8m
5 km
15 m
7 km
.10 km
41.5
m/s
11 m
0.8 m/s
2.5 m
2.5 m
41.5
m/s
11.7 m
0.8 m/s
3m
1m
41.5
m/s
15.4 m
1.2 m/s
2.5 m
0.75 m
11 m
20 km
54 - 70
km
43
m/s
8.1 m
1.0 m/s
3m
0.75 m
40.5
m/s
6.41 m
0.4 m/s
2.85 m
0m
43
m/s
10.1 m
0.4 m/s
2m
0m
1.25 m
1.2 m
1.4 m
1.6 m
0.5 m†
0.64 m†
1.21
1.23
1.23
1.35
1.23
1.23
0
0
0
0
0.6m
0.4 m
gravelly
sand
medium
dense
sand
medium
dense
sand,
silted
medium
to dense
sand,
gravel,
boulders
Wave data at SE-1 from measurements, data for DE-1 related to SE-1
Table 10.5: Compatibility of support structure concepts and sites
Concept
UK-1
NL-1
NL-5
DK-1
DE-1
SE-1
Monopile
(soft-stiff or soft-soft †)
+
++
-
++
++
+
Gravity monotower
(soft-stiff)
+
+
o
o
-
+
Lattice tower
(soft-stiff, piled or gravity)
-
-
o
-
--
--
Tripod
(soft-stiff, piled or gravity)
+
+
++
o
--
-
†
soft-soft dynamic characteristics only promising for rotor speed above 19 rpm
229
500
6.5
Project
management
400
6
Wind turbine
300
5.5
200
5
100
4.5
Support
structure
UK-1
NL-5
DE-1
soft-soft
NL-1
soft-soft
DE-1
SE-1
NL-1
Grid
connection
Levelised
production
cost
4
DK-1
0
Energy cost [€ct/kWh]
Captial cost [M€]
6.5
1000
6
750
5.5
500
5
250
4.5
UK-1
NL-5
DE-1
soft-soft
NL-1
soft-soft
DE-1
SE-1
4
NL-1
0
Energy costs [€ct/kWh]
1250
DK-1
Annual energy yield [GWh/a]
Figure 10.3: Capital cost and energy cost of preliminary designs
Gross energy
yield
Net annual
energy yield
Levelised
production
cost
Figure 10.4: Energy yield and energy cost of preliminary designs
60
900
58
pile
related cost
800
56
hub height
700
54
600
52
tower related cost
500
Hub height (MSL) [m]
Pile, tower cost [k€]
1000
50
5
10
15
Water depth (MSL) [m]
20
25
Figure 10.5: Capital cost for tower and pile including OWEC installation
(left axis) and hub height (right axis) against water depth
230
Chapter 10
Similar NL-5 with the prime wind conditions suffers from its remote location leading to
higher dielectric transmission losses and lower availability in addition to the high
investment for the shore connection and support structure. A cost level competitive to
the best two sites would require lower investment and higher energy yield, e.g. due to
a taller structure. This could be achieved, for instance, by 30% lower investment for the
support structure and 12% more energy, corresponding to an increase of the hub height
from 60 m to 100 m. It is unlikely that a tripod design can manage this task alone. More
mature technology with higher efficiency and/or higher power will be required.
Now we spend a closer look to the support structure properties in Table 10.6. Generally
the stiffness criterion and the aerodynamic or combined wind and wave fatigue
determine the design. The tower base diameter is chosen equal or smaller than the
subsoil pile diameters. Therefore, it is not possible to push the tower design into buckling
determined conditions. The design eigenfrequency of all designs is very close to the
exclusion range at the rotor resonance. Large weights and pile dimensions make the
soft-soft NL-1 and the soft-stiff NL-5 not feasible. The other concepts are regarded as
reasonable preliminary designs.
For all soft-stiff concepts the minimum allowed hub height with respect to wave and
blade clearance, (10.2), is optimal. Only at the shallow site DE-1 a soft-soft monopile
with a double conical tower and 80 m hub height is realistic and results in approximately
the same energy cost as a shorter soft-stiff solution.
Figure 10.5 plots the pile and tower investment including cost for site investigation and
OWEC installation of the soft-stiff designs against the water depth. Both pile and tower
costs are increasing proportionally to the water depth. Together the increase is
approximately 5% per metre water depth leading to a 60% more expensive structure at
the British site with 20 m depth compared to the German site with only 8 m.
The joint between pile and tower is located at the platform level. This means that for
larger water depth the increase in hub height results in a longer pile and in a constant
tower length since always the minimum hub height is chosen. One might be wondering
why the tower cost are still rising with water depth. The reasons lie in the higher wind
Table 10.6: Main data of preliminary monopile support structures
†
DE-1
(NL-1
softsoft †)
DE-1
softsoft
(NL-5 †)
UK-1
56
54
107
80
60
59
0.40
0.39
0.40
0.26
0.26
0.40
0.40
3.75
4.29
3.8
3.5
5.25
3.5
5.1
4.5
Pile wall
thickness [mm]
75
75
75
70
105
85
80
80
Tower mass [t]
99
98
107
91
(393)
234
(106)
101
Pile mass ‡ [t]
311
432
357
248
(744)
309
(614)
497
DK-1
NL-1
softstiff
SE-1
Hub height
above MSL [m]
56
58
Eigenfrequency
[Hz]
0.40
Pile diameter [m]
Monopile design not viable under this conditions.
‡
including transition piece
231
loads for more exposed sites and especially the extra material needed to maintain the
soft-stiff characteristics for a taller structure.
Selection of site and concepts for further development
The conceptual design should consider either the DK-1 or NL-1 site. The 5% difference
in energy cost between both sites is not significant with respect to the accuracy of the
estimates. Moreover the onshore grid connection cost at the Danish site is considerably
higher and would more than compensate the price difference since a 150 kV connection
point for a maximum 250 MW is 54 - 70 km away. A 400 kV grid node is even 150 km
further lacking [10.18]. In contrast, heavy industry and an oil fired plant are located at
IJmuiden, the shore connection point for NL-1 [10.1].
The availability of a met-ocean database is another scientific argument supporting the
Dutch NL-1 site.
Both the soft-stiff and the soft-soft monopile offer opportunities. The conceptual design
phase should trace a wind turbine - structure configuration that is better compatible with
the high aerodynamic fatigue loads. In particular, a higher rotor speed, which would also
result in cost reductions of the wind turbine, should be considered.
10.2 Conceptual design
During the conceptual design the specific properties of the selected sites and concepts
are considered and the principal dimensions of the designs are determined. Based upon
criteria related to the entire wind farm, in our case the levelised production cost, the final
design and site are chosen.
10.2.1
Treatment of dynamics and design
Large offshore wind farms should be designed and certified according to the local site
conditions which establish the base for a so-called ‘S-Class Certification’ according to
the Germanischer Lloyd or IEC 61400-1 ed. 2 standard.
If no site measurements are available, dedicated computer models or met-ocean
databases have to be evaluated by consultants. In the best case a three-dimensional
scatter diagram with classes for significant wave height, Hs , zero-crossing period, Tz ,
and mean wind speed at hub height, V, is available. Otherwise a two-dimensional wave
scatter diagram has to be correlated with the Weibull wind speed distribution by
comparing the cumulative distributions (Section 8.2).
If no joint-probabilities for the extreme wind, wave and current conditions exist, the
individual extremes have to be combined according to the Germanischer Lloyd’s
guidelines.
Again wind and waves can be considered as collinear and omnidirectional.
A finite element model of the support structure should be used for the determination of
the relevant eigenfrequencies. Evaluation of a Campbell diagram [10.19] reveals
possible resonances between natural frequencies and rotor excitations.
232
Chapter 10
Depending on the available design tools and data, the soil stiffness of piled foundation
can be either modelled by distributed springs with P-y curves or equivalent foundation
springs for the translational and rotational degrees of freedom. Variations in water depth
and soil conditions should be accounted for.
Aerodynamic damping
The variation of the aerodynamic damping for different support structure concepts and
for different mean wind speeds should be studied. Especially for soft designs, potentially
prone to hydrodynamic fatigue, a detailed description is required for the determination
of the hydrodynamic response.
Different methods for the determination of the damping are discussed in Section 7.1.3.
Attention should be given to the possible reduction of the damping ratio at post-rated
wind speeds due to stall or the dynamics of the control system.
Once soil parameters are available one can investigate the sensitivity of the foundation
behaviour with respect to variations in the secondary soil parameters, e.g. the stiffness
of the entire soil or of certain soil layers. Variation of the primary soil parameters as the
internal friction angle or shear strength is most worthwhile when these parameters have
become available from soil investigations. Therefore, such a treatment is proposed at
least for the detailed design phase.
Because foundations tend to behave stiffer than expected, the positive tolerance in the
soil parameters may be taken as greater compared to the negative tolerance. In the
sensitivity study of the reference designs in Section 9.5.2 indicative safety factors of γm
= 0.5 and γm = 3.0 have been considered uniformly for all soil layers. After such analysis
the design ranges for the eigenfrequencies can be established on a more rational basis
than in the previous section.
Critical load cases should be identified for the different design concepts. Apart from
extreme load cases mentioned in Section 10.1.1 other conditions, e.g. emergency stop,
(near-)breaking waves, breaking ice, might be of relevance.
Superposition of aerodynamic and hydrodynamic extreme responses from separate
computations is reasonable.
The aerodynamic and hydrodynamic fatigue can be calculated with the simplified
analysis based upon the superposition of short-term fatigue (Section 8.1) if correlated
wind and wave data exist.
Economic evaluation
An OWECS cost model can be of great value also during the conceptual design. While
the model might be still used for the overall evaluation, the cost of certain subsystems
can be estimated directly by input from the designer or from suppliers rather than by the
general cost algorithms of the model.
10.2.2 Conceptual design of the Opti-OWECS wind farm
During the conceptual design phase of the Opti-OWECS study a simultaneous
optimisation of the wind turbine and of the support structure was carried out. The main
goal was to reduce cost by lowering the quite considerable aerodynamic fatigue loads.
Therefore a parameter study on aerodynamic fatigue loads as function of rotor speed
233
and support structure stiffness was carried out by time domain simulations of about 20
different configurations. It was not practical to consider combined wind and wave fatigue.
This was a major limitation because softer designs were subject to lower aerodynamic
but higher hydrodynamic fatigue. The final Opti-OWECS design combined a higher rotor
speed of 22 rpm instead of 19 rpm and a softer monopile structure with a natural
frequency of 0.29 Hz instead of the soft-stiff gravity monotower with 0.52 Hz considered
as conceptual Opti-OWECS design. The solution was found only after a time-consuming
iterative design process including several sets of time domain simulations of the
combined response. Nonetheless an approximately 50% (!) reduction of the
dimensioning loading was achieved.
In the scope of this chapter the parameter study is repeated by the author. Now it is
possible to account for combined fatigue by one of the simplified fatigue analysis
methods. For this study with many different configurations the long-term superposition
of the aerodynamic and hydrodynamic fatigue (Section 8.2) rather than the abovementioned superposition of short-term fatigue responses is chosen.
Comparison of rotor design options
In the scope of the Opti-OWECS study, van Rooij [10.1] investigated to reduce the
extreme parking and fatigue loads and blade weight by maintaining a reasonable energy
yield for constant diameter and rated power. The exercise started with the rotor design
of the land-based version of the WTS 80 with 19 rpm. Two higher rotor speeds of 22 and
25 rpm were considered for an annual average wind speed of 8.5 m/s at hub height. The
higher rotor speeds result in tip speeds of 92 and 105 m/s which are only feasible due
to the relaxed noise constraints offshore. In general, a larger rotor speed decreases the
required chord length along the blade span.
Enlarging the rotor speed from 19 rpm to 22 rpm the chord reduction was found to be
significant even for a constant maximum chord at the blade root and same airfoils. The
new blade design for 22 rpm (Figure 10.6) obtains approximately 1% higher energy yield
than the original design. A further reduction of the blade area for rotating at a higher
speed of 25 rpm results in a slight change of the maximum power coefficient but the
effect on the total energy yield is small. Hence no adjustment of the blade geometry was
proposed for 25 rpm.
The result should be seen as the first step towards a new rotor design for offshore
application. The weight of the blade and thus also the cost will be reduced for the new
chord distribution and the more fatigue benign offshore wind condition. A structural
re-design of the internal blade structure was not further investigated since it would
require a major involvement of a blade manufacturer.
Comparison of support structure configurations
A fictitious monopile support structure is considered for the parameter study. The
stiffness of the reference configuration with f0 = 0.4 Hz is varied by artificially adjusting
the modulus of elasticity by a factor between 0.4 and 3 corresponding to a first
eigenfrequency between 0.25 and 0.69 Hz. This approach works well for the calculation
of the aerodynamic tower top loads, however, results in simplifications concerning the
contribution of the inertia loading along the support structure. The damage equivalent
tower top loads are generated by time domain simulations for various combinations of
stiffness and rotor speed whilst spectral analysis is applied for the hydrodynamic fatigue
loads at the NL-1 site with 15 m water depth (Table 8.1). Direct quadratic superposition
obtains the combined fatigue loads.
234
Chapter 10
Chord [m]
5.00
4.00
Original blade: 19 rpm
3.00
2.00
Optimised blade: 22 rpm
1.00
Thickness [%]
Twist [deg]
0.00
15
12.5
10
7.5
5
2.5
0
0
0
5
5
10
10
15
15
20
20
25
25
30
35
30
35
30
35
40
40
50%
40%
30%
20%
10%
0%
0
5
10
15
20
25
40
Spanwise radius [m]
Figure 10.6: Original blade layout of WTS 80 for rotor speed 19 rpm
and optimised blade for 22 rpm
Figures 10.7 to 10.9 show the equivalent bending moment of the monopile 8 m below
the mudline as function of the support structure stiffness and for different loadings.
Aerodynamic fatigue loads for a rotor speed of 22 rpm (solid line in Figure 10.7) increase
rapidly in the vicinity of the strong 2P blade excitation. The quite pronounced
amplification of the aerodynamic fatigue loads in the vicinity of the rotor frequency (1P)
originates from the mass unbalance of the rotor according to the Guidelines of
Germanischer Lloyd. In practice the effect might be considerably lower because the
eccentricity of recent wind turbine rotors is 5 to 10 times lower.
Softer configurations exhibit higher hydrodynamic fatigue. The increase of the dynamic
amplification is, however, partially reduced due to the higher aerodynamic damping ratio.
For the stiffer structures the loading lowers decreasingly towards the quasi-static
response (dotted line in Figure 10.7). The ratio between the wave fatigue loading of
different pile diameters (dashed lines) is proportional to the square of the diameter owing
to the dominant hydrodynamic inertia forces.
In general, wind generated fatigue dominate the investigated configurations. Only for
very soft structures and large pile diameters the hydrodynamic fatigue is more important
than aerodynamic fatigue.
Figure 10.8 combines the fatigue loads for the two pile diameters from the previous
figure and shows the pure aerodynamic loads for a rotor speed of 22 rpm as reference.
The quadratic superposition of the load components is clearly visible. A soft-soft
solution is beneficial only for a pile diameter which is small, at least close to the water
surface, while a soft-stiff design is reasonable also for larger piles. Hence, Figure 10.9
235
wind
waves ( 4 m pile )
40
waves ( 3 m pile )
1P resonance
30
20
waves (3 m, quasi-static)
2P resoance
equivelant bending moment [MNm]
50
10
0
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0.6
fundamental eigenfrequency [Hz]
0.65
0.7
Figure 10.7: Separate wind and wave fatigue loading against support structure
stiffness for rotor speed 22 rpm
40
30
2P resoance
1P resonance
equivelant bending moment [MNm]
50
20
wind & waves ( 4 m pile)
10
wind & waves ( 3 m pile)
wind
0
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
Figure 10.8: Combined wind and wave fatigue loading against structure stiffness for
rotor speed 22 rpm
25 rpm
equiv. bending moment [MNm]
70
60
Opti-OWECS
design solution
19 rpm
22 rpm
50
40
30
Opti-OWECS
conceptual design
20
10
0
0.25
Soft-soft
conceptual design
0.3
0.35
3 m pile
4 m pile
Soft-stiff
conceptual design
0.4
0.45
0.5
0.55
0.6
0.65
Figure 10.9: Combined wind and wave fatigue loading as function of support
structure stiffness and rotor speed
236
Chapter 10
presents the combined fatigue for a 3 m pile in the soft-soft region (solid line) and for a
4 m pile in the soft-stiff range (dashed lines). Comparison of the three rotor speed
variants confirms the prospects of both a soft-soft solution with either 22 or 25 rpm and
a soft-stiff design with higher total loads for the original rotor speed of 19 rpm.
Support structure concepts
With aid of the MonOpt code two monopile support structures are designed with a hub
height of 58 m (MSL). The water depth corresponds to the maximum value of the NL-1
site ranging between 15 and 20 m (MSL).
The modelling from the previous section is mainly followed and is not repeated here
again except a remark on the design ranges for the fundamental eigenfrequency. The
±15% tolerance for structural uncertainty from the feasibility study is maintained in order
to anticipate on the variation in water depth to be considered in the detailed design
phase.
The designs are similar to the Opti-OWECS solution (Figure 9.4) with a tapered section
below the water surface and a grouted joint between pile and transition piece. The
reduction of the water piercing cross section is absolutely required to keep the
hydrodynamic loads and required amount of material within an acceptable range.
Figures 10.10 and 10.11 compare the geometry and the strength utilisation of the softstiff and the soft-soft concept. Both designs are driven by fatigue and stiffness
requirements. The fatigue strength utilisation along the height is given in the lowest of
the subplots. In addition, the buckling utilisation of the soft-stiff design and the
contribution of hydrodynamic fatigue to the total utilisation at the soft-soft design are
shown.
Both designs are feasible but show also potential problems. Series installation of the
soft-stiff monopile with the diameter of 4.7 m and enormous weight of over 350 t is
challenging and beyond the capabilities of most contractors. The soft-soft pile can be
easier and faster handled but the large wall thickness at the grouted joint is difficult to
manufacture. So, the performed cost estimates, mainly based upon the steel weight, are
only indicative.
Evaluation on global criteria
The decision on the final design, the WTS 80M turbine with 22 rpm and a soft-soft
monopile is made after a trade-off between the wind turbine and support structure
options (Table 10.7). The higher rotor speed yields lower fatigue loads and enables a
weight reduction of the rotor blade and the support structure. Extreme operating loads
are likely to be higher for this options but in total, fatigue is governing. A saving in the
wind turbine investment of 5% is conservatively estimated. In addition, the new blade
gains slightly more energy. The soft-soft monopile offers further significant cost
reduction. In total, the energy cost is estimated to be approximately 5% lower than for
the base case.
The achieved saving in this section is lower than during the Opti-OWECS project where
a soft-stiff gravity monotower with the WTS 80 was compared to the soft-soft monopile
with the WTS 80M (Figure 10.9). Thanks to the integrated methodology the relatively stiff
gravity monotower option has already been discarded during the feasibility study in
Section 10.1.2 and only an evaluation of two relatively similar monopile designs has
been made.
237
2
1
Thickness [mm]
0
-50
-40
-30
-20
Flange
3
Grouted
Joint
4
Mudline
Diameter [m]
5
-10
0
10
20
30
40
50
-10
0
10
20
30
40
50
60
125
100
75
50
25
0
-50
-40
-30
-20
60
Utilisation [%]
100%
Fatigue
75%
50%
Buckling
25%
0%
-50
-40
-30
-20
-10
0
10
20
30
40
50
30
40
50
30
40
60
Figure 10.10: Soft-stiff monopile concept
(Pile 361 t, transition piece 90 t, tower 101 t)
2
1
Thickness [mm]
0
-50
-40
-30
-20
-10
-20
-10
Pile incl. transition piece
Flange
3
Grouted
Joint
4
Mudline
Diameter [m]
5
0
10
20
0
10
20
Tower
60
125
100
75
50
25
0
-50
-40
-30
50
60
Utilisation [%]
100%
75%
Combined Fatigue
50%
25%
Hydrodynamic Fatigue
0%
-50
-40
-30
-20
-10
0
10
20
30
Figure 10.11: Soft-soft monopile concept
(Pile 300 t, transition piece 94 t, tower 98 t)
40
50
60
238
Chapter 10
Table 10.7: Comparison of conceptual designs for NL-1 at 20 m (MSL)
WTS 80 (19 rpm)
soft-stiff monopile
WTS 80M (22 rpm)
soft-soft monopile
Turbine investment
0%
-5%
Support structure investment &
OWEC installation
0%
-8%
Energy yield
0%
+1.3%
Levelised production cost
0%
-5%
10.3
Detailed design
During the detailed design the final concept is worked out and drawings, specification
and the bill of material are produced. In our case the support structure design is
specified for the entire wind farm including variation of the local parameters.
The environmental description should be based upon correlated wind and wave data
obtained for the actual site from measurements or hindcasted data (Section 4.2).
Consideration of directional data, as shown in Figure 4.17, is desirable because of the
large differences in fetch between the landwards and the seawards directions.
Integrated non-linear time domain simulation should be used for the final load
calculations. So, lumped sea states have to be established for the fatigue calculations,
as described in Section 8.3, in order to limit the computational effort to a reasonable
extent.
The sensitivity of the eigenfrequencies and the damping with respect to a systematic
variation of all environmental, structural and operational properties should be studied
with a finite element code and/or an integrated OWEC design tool. Tables 5.5 and 5.6
list conditions resulting in either minimum or maximum eigenfrequencies and damping.
Beforehand it is not always clear whether the highest aerodynamic and hydrodynamic
loads will occur for the minimum or the maximum of the expected eigenfrequencies. For
fatigue loads which accumulate over long time it might be appropriate to assume
average conditions, e.g. mean sea level, moderate marine growth, low lateral pile
loading, etc. So different structural models of the OWEC should be established for
fatigue and extreme load calculations assuming the most onerous conditions.
Aerodynamic damping
The application of integrated, non-linear time domain simulations ensures the most
accurate description of the aerodynamic damping. In particular the effects of stall and
of the dynamics of the control system are fully taken into account.
239
The most rational approach for estimating the foundation uncertainty is to investigate the
sensitivity in the primary soil parameters directly. Taking sand as example, these are the
internal friction angle, the modulus of subgrade reaction and undrained density. The
behaviour of the entire OWEC is studied also with respect to other uncertainties as
scour and settlement. The location and the magnitude of the extreme stresses in the
foundation vary for different soil conditions. Analysis of the foundation uncertainty is one
part of the above-mentioned systematic investigation of the variation of the dynamic
characteristics.
If combined aerodynamic and hydrodynamic fatigue is important, it should be analysed
by non-linear, time domain simulations of lumped load cases (Section 8.3). Otherwise,
one of the simplified analysis approaches might be sufficient for a check of the design.
In case extreme loads govern the design, other dedicated analysis methods such as
structural reliability methods [10.20, 10.21] or push-over analysis with a non-linear finite
element code might be adequate.
Economic evaluation
Investment and energy costs could be calculated on the basis of the estimated
component costs. Uncertainties in the gross energy yield, OWECS availability, operation
and maintenance cost, investment cost and economic parameters should be evaluated
to establish the most likely energy cost and its uncertainty.
10.3.2 Detailed design for the Opti-OWECS wind farm
Variation of soil conditions
In absence of actual soil investigations for different locations of the NL-1 site we assume
medium dense sand within the entire wind farm. Referring to Table 4.6 the uncertainty
of the soil parameters is established for the subsequent dynamic simulations (Table
10.8). The initial subgrade reaction is varied much larger compared to the values
recommended by the API Guidelines [10.22] for a friction angle of 29o and 36o,
respectively. For the design calculations the minimum soil stiffness is assumed.
In an additional parameter study the sensitivity of the first two fore-aft eigenfrequencies
at 20 m water depth (MSL) is investigated for a wider range of soil properties than
expected at the site (Figure 10.12). Now the relation between friction angle and
Table 10.8: Variation of design soil conditions
Comment
Submerged
unit weight γsub
[kN/m3]
Internal friction
angle N
[deg]
Initial
subgrade
reaction K
[MPa/m]
Minimum
soil stiffness
8
30
7
cylic loading, preloading 300 kN
Maximum
soil stiffness
10
36
40
static loading,
no pre-loading
Design
condition
sensitivity [%]
240
Chapter 10
5.0%
0.0%
P-y curve static loading, no pre-loading
-5.0%
sensitivity [%]
-10.0%
P-y curve cyclic loading, with pre-loading
1st fore-aft natural frequency
27.5
30
32.5
35
37.5
40
42.5
45
10%
0%
P-y curve static loading, no pre-loading
-10%
-20%
P-y curve cyclic loading, with pre-loading
2nd fore-aft natural frequency
27.5
30
29 30
3.5 8.5
32.5
35
37.5
internal friction angle Phi [deg]
35
40
42.5
45
40
45
43
23
initial subgrade reaction K [MPa/m]
63
Figure 10.12: Sensitivity of the fore-aft natural frequencies of the NL1-20m
monopile to simultaneous variation of the internal friction angle and the initial
subgrade reaction at a submerged unit weight of 10 kN/m3
subgrade reaction according to API is applied (see lower axis). Acceptable behaviour
is observed for medium dense to very dense sand but the frequencies drops
dramatically if only loose sand or medium dense sand with portions of silt and clay
occur. Under such conditions the structure becomes too soft and is no longer viable as
consequence of excessive wave excitation. Further reinforcement of the wall thickness
is not practicable and an increase of the outer diameter is more than absorbed by the
higher (inertia dominated) wave loading proportional to the square of the structural
dimension. In such a situation one has to choose a tripod or gravity monotower as
alternative.
The influence of scour is not considered since some form of scour protection is
generally required in the Southern North Sea. For this structure with pronounced
dynamic behaviour the avoidance of seabed erosion is essential.
Analysis of combined wind and wave fatigue
The soft-soft monopile design is driven by fatigue. Therefore, time domain simulations
of the simultaneous wind and wave response with lumped load cases from Table 8.3 are
performed. The response during idle or failure state, when no aerodynamic damping is
present, is more onerous than the combined aerodynamic and hydrodynamic fatigue
during production. As illustration Figure 10.13 shows the fatigue loading along the
structure for both 85% and 100% availability. Over the entire length the former
conditions results in considerably higher loads.
At the grouted joint and below the seabed the utilisation for omnidirectional loading even
exceeds slightly 100%. A proper consideration of the different loading for different
directions would require several times higher computational efforts. In a simplified
241
equivalent bending moment
[MNm]
40
35
30
25
20
15
Fatigue resistance
10
Fatigue demand (85% availability)
5
Fatigue demand (100% availability)
0
-30
-20
-10
0
10
20
30
40
50
Figure 10.13: Fatigue loading of the monopile at 20 m water depth for
different availabilities (Omnidirectional and collinear wind and wave loading)
approach the effect of the directionality of the loading is estimated. The omnidirectional
response is distributed over eight wave direction sectors with the frequency from Figure
4.17. A priori it cannot be judged whether this treatment is conservative, probably it is
not, therefore it is used only qualitatively. The damage equivalent stresses are reduced
by about 20% and the estimated life is increased by 120 to 140%. The minimum
calculated fatigue life occurring at the grouted joint is estimated to be between 17 and
39 years; most likely it is closer to the upper bound of this range and therefore larger
than the design life of 20 years.
Local site conditions
So far we have considered only the maximum water depth in order to check the viability
of the design concept. During the detailed design the variation of the mean sea level
between 15 and 20 m and different soil conditions need attention. We assume that 100
offshore wind energy converters are uniformly distributed over the range of local
conditions and that the same met-ocean data are valid all over the site.
Adaptation of the support structure to such local conditions should balance logistic and
procurement efforts against the possible cost saving owing to tailored design. After a
parameter study it is decided to propose two monopile designs suited for water depths
of either below or above 17.5 m. The same tower and transition piece can be used,
however, the pile length and stiffness varies. With respect to the pile driving equipment
it is advantageous to maintain the pile diameter and to restrict the modifications to the
pile length and wall thickness. Furthermore a lower wall thickness eases the
manufacturing. Both piles have constant length above MSL and below the mudline, as
the height of the cylindrical part between the seabed and the tapered section is adjusted
to the local water depth.
Table 10.9 compares the main properties and the indicative costs of both designs,
denoted as NL1-17.5 and NL1-20. The different water depth and soil conditions result
in an approximately 8% to 10% spreading of the first natural frequency. Under all
circumstances at least 10% distance to the 1P rotor excitations remains.
242
Chapter 10
A further weight reduction of the monopile might be possible if a reduced wall thickness
of, say, 40 mm at the lower half of the penetration depth where pile loads are low and
the soil stiffness is high. Such measures require, however, a check of both possible
buckling during installation and increased hydrodynamic excitation.
The structural modifications affect the energy cost 7 only by a fraction. Nonetheless they
are required in order to avoid a too stiff characteristics at low water depths.
Table 10.9: Final monopile designs for different water depth
Design
Property
Water depth (MSL)
Hub height (MSL)
1st fore-aft natural frequency †
Pile dimensions
NL1-17.5
NL1-20
15 - 17.5 m
17.5 - 20 m
58 m
58 m
0.292 - 0.33 Hz
(0.80 - 0.90 P)
0.297 - 0.323 Hz
(0.81 - 0.88 P)
Ø 3.65 m x 60 mm
x 40 - 42.5 m
Ø 3.65 m x 70 mm
x 42.5 - 45 m
Tower mass
98 t (identical design)
Transition piece mass
92 t (identical design)
Pile mass
225 - 239 t
272 - 287 t
Investment for support structure and OWEC
installation
. 1.12 Mi
. 1.19 Mi
Levelised production cost
†
. 4.4 iCt/kWh
. 4.5 iCt/kWh
including variation of water depth and uncertainty of soil conditions
General remarks
The example within this chapter might be driven by the specific wind turbine design.
Nonetheless, the demonstrated methodologies for the optimisation of the dynamics of
both wind turbine and support structure and the adaptation to the site parameters is
promising for different types of bottom-mounted offshore wind energy converters.
Most wind turbine manufacturers design standard machines and major modifications of
these so-called ‘platforms’ to tailor them to certain wind turbine classes are limited
(Section 3.2). Therefore a project specific design of wind turbine platforms is
questionable even for larger offshore projects. Nonetheless, support structures, grid
connection and procedures for installation, operation and maintenance will differ from
project to project and an integrated design approach is certainly beneficial. Major
parameters to be adjusted will include: foundation type, height of the access platform,
hub height, dynamic characteristics, tower design, wind turbine control features and
possibly small variations of rotor speed, diameter or generator power.
7
The average energy cost of 4.5 iCt/kWh at an annual average wind speed of 9 m/s at hub height
corresponds to the estimate of 5.1 iCt/kWh for 8.4 m/s annual average wind speed of the Opti-OWECS
project. After the European project it turned out that the reference height in the NESS database was
accidentally taken as too high, which caused a too low estimate of the energy yield.
Conclusions
243
CHAPTER 11
CONCLUSIONS
The excellent prospects of offshore wind energy depend, from a technical perspective
firstly on further reduction of energy cost through innovative design and economies of
scale, and secondly upon the integration of large quantities of wind energy generated
electricity into the international energy system.
The work described in this thesis contributed to the former aspect. More specific
conclusions are related to the three major fields of the research. In addition, some
recommendations for further work are given.
Design approach for offshore wind farms
The economic exploitation of the offshore wind energy potential requires a symbiosis of
wind energy, offshore technology and power management. Although this seems
obvious, progress is by no means evident and requires an open attitude in each
community.
Design methodologies recently developed for complex civil engineering systems are
also adequate for offshore wind farms. Within the novel, integrated design approach for
offshore wind energy conversion systems the goals of the design process are controlled
effectively on the system level by overall criteria such as levelised energy costs,
adaptation to site conditions, installation and commissioning efforts, availability and
overall dynamics.
With respect to the economies of scale and the required development time a project
specific design of wind turbine platforms is questionable even for larger offshore
projects. Nonetheless, support structure, grid connection and procedures for installation,
operation and maintenance will differ from project to project and an integrated design
approach is beneficial.
An integrated approach is also required when designing wind turbines for the required
high reliability and availability at exposed offshore sites.
The application of a common offshore wind energy terminology is valuable and more
than pure convention.
A separate handling of tower and foundation might be more convenient in the short-term,
but probably only a sub-optimum design solution will be achieved compared to the
treatment of the support structure as one part. Likewise, the consideration of the wind
turbine and support structure as one entity, the offshore wind energy converter, results
in a potentially more appropriate and more economic designs.
Design of an offshore wind energy conversion system includes not only the wind
turbines and their support structures but likewise the operations and maintenance
aspects and the grid connection.
244
Chapter 11
Dynamics of offshore wind energy converters
An integrated, non-linear dynamic model of an offshore wind energy converter (OWEC)
in the time domain accounts for the interactions between the subsystems and the
significant non-linearities in the dynamics and the aerodynamics of the wind turbine. It
is required for certification calculations and analysis of the non-linear response under
extreme conditions. The high computational effort for the analysis of the simultaneous
aerodynamic and hydrodynamic response under fatigue conditions can be reduced
using a new method which constructs a low number of lumped load cases.
For the early design stage, two simplified analysis methods were developed. They are
based upon the superposition of the damage equivalent stress ranges from separate
analyses of the aerodynamic and hydrodynamic response with standard design tools in
the time domain and frequency domain, respectively. The weighted quadratic
superposition is founded on a theoretical base but showed larger scatter than the direct
quadratic superposition of damage equivalent loads which was established empirically.
The approaches require the estimation of the aerodynamic damping of the support
structure. A generally available method is the evaluation of transient vibrations in steady
wind for a turbine with constant rotor speed or in turbulent wind for variable speed
turbines. As a conservative measure only 80% of the estimated damping should be
applied as additional structural damping of the fundamental fore-aft mode when
analysing the hydrodynamic response alone.
One particular reference design of a pitch regulated, constant-speed offshore wind
energy converter with a soft-soft support structure has been investigated with the novel
methods. The results show good agreement with integrated, non-linear time domain
simulation and suggest that an accuracy of ±10% or better in terms of long-term
equivalent loads compared to simulations could be achievable by proper application.
Further studies at different configurations are required to judge a general accuracy.
The third, new method of constructing a low number of characteristic load cases by
lumping of sea states has been demonstrated as an accurate and effective method to
reduce the computational efforts for time domain simulations to a reasonable extent.
In the investigations of the structural dynamics of offshore wind energy converters the
support structures was emphasised because it is most specific and most affected by the
offshore siting. Its dynamics are largely determined by the combination of:
• first fore-aft and lateral natural frequency,
• associated mode shape and global distribution of support structure stiffness,
• response of the second and higher eigenmodes, which can be important for
extreme loading associated with breaking waves and dynamic ice loading,
• magnitude of aerodynamic damping ratio, depending on wind turbine and support
structure design and operational conditions,
• sensitivity to loading, e.g. water-piercing cross section of the structure, ice cone
• site specific environmental parameters. (High quality site data are absolutely
required for proper support structure design, while the wind turbine could be
designed with a generic classification of conditions compatible with the actual site.)
• wind turbine characteristics including control.
Understanding is far from easy but the variety of influences offers opportunities to tailor
the dynamics and to optimise the design.
Conclusions
245
A relatively stiff foundation together with a soft tower lowers the sensitivity to the soil
conditions, reduces the dynamic wave loading and can improve the economics. Such
an integrally designed structure contradicts the rule-of-thumb from the offshore
engineering which says that structures with a natural frequency below 0.4 Hz should be
avoided with respect to dynamic wave excitation.
Although lower support structure stiffness can be beneficial with respect to the wind
turbine induced fatigue loading, it attracts hydrodynamic fatigue and can cause high
extreme loads during emergency shut-down or extreme wind conditions.
High support structure stiffness may conflict with wind turbine induced fatigue and
generates high structural and installation costs.
At demanding offshore sites, the hydrodynamic fatigue loading during parking, when no
aerodynamic damping is present, can exceed the combined aerodynamic and
hydrodynamic fatigue during power production. Consequently, fatigue analyses have to
consider different values of availability.
Aerodynamic damping is the most significant source of damping while hydrodynamic
damping is negligible for bottom-mounted structures and soil damping of the
fundamental mode of piled or gravity systems is low.
Dynamic analyses have to account for the uncertainties and variability in the soil
properties and met-ocean parameters. Suitable treatments for different stages of the
design process have been proposed. Large safety factors should be applied on
estimated soil parameters even after in-situ tests and both lower and upper bounds
require attention.
Most analyses were carried out for a 3 MW, 80 m two-bladed wind turbine with constant
speed - variable pitch control. This uncommon machine concept generates high fatigue
loads and does not fully represent the multi-megawatt machines currently under
development. The use of three blades, larger rotor diameters, variable rotor speed and
advanced control affect lower specific loads and may result in distinct dynamic
behaviour. Nonetheless, there are indications that fatigue becomes also the design
driver for more advanced multi-megawatt turbines with a lower specific rating, i.e. larger
rotor diameters. In such a case the presented results are valid in a qualitative way.
It is expected that the monopile concept is technically and economic viable only up to
a water depth of 15 - 20 m (MSL). Up to such water depths, for moderate pile diameters
close to the water surface and not too soft structures, the wind or ice loads are higher
than wave and current loads. Beyond such conditions the required diameter and/or the
wall thickness become too large with respect to installation and construction,
respectively.
Tripod support structures are promising for greater water depths and larger turbines. The
high stress concentrations at the tubular joints are an inherent disadvantage which is
conflicting with the significant magnitude of the fatigue loads for multi-megawatt turbines.
Design optimisation of offshore wind farms
A site specific design and certification is economical attractive for offshore wind farms.
The higher mean and lower dynamic wind loads and the very specific hydrodynamic and
geotechnical conditions are not compatible with standardised onshore designs.
In comparison to the land-based situation, the relaxed noise restrictions, importance of
reliability and operation and maintenance aspects and higher installation cost favour
eventually distinct wind turbine design for offshore siting.
246
Chapter 11
Dedicated offshore design can be more reliable and safer. The example of the wind farm
Lely was used to show how the sensitivity to variations in the soil properties and the
fatigue due to wind and wave response can be reduced leading to a more durable
design solution.
Design optimisation by means of tailored dynamics can also achieve lower capital and
energy costs as illustrated during the Opti-OWECS study, by the development of a 300
MW wind farm in the Dutch North Sea through three successive stages of the design
process.
Recommendations for further research and development
First experience with the integrated OWECS design approach was gained during the
Opti-OWECS project. Obviously such a methodology needs further development by
usage in desk-top studies as well as implementation projects.
Under the conditions and constraints of large ‘real’ projects, it has to be picked out in
which form the approach is suitable and compatible with the different organisations
involved.
The proposed Design for RAMS (Reliability, Availability, Maintainability and
Serviceability) design approach has not been investigated in great depth so far and its
procedural as well as scientific development is a worthwhile project of its own.
Establishment and harmonisation of specific standards for offshore wind farms have not
progressed very far, yet. The existing guidelines of Germanischer Lloyd and the Design
Bases for Offshore Wind Farms in Denmark should be further developed and validated
through measurements. Such work has started in the scope of two EU projects and
national projects, recently and should be incorporated in the compilation of a separate
offshore extension of the international IEC 61400 series.
In the same scope the time domain analysis approach of the overall dynamics should
be further refined and verified by measurements. The DUWECS simulation code needs
further develoment. Examples include the aerodynamic and structural modelling of the
wind turbine and the tower-nacelle interface.
The risk involved in large offshore wind farms could be reduced by improving the
understanding of the offshore wind conditions and associated other met-ocean
parameters in order to facilitate more reliable predictions of the energy yield, optimum
design and optimum strategy for installation as well as operation and maintenance. The
use of high quality met-ocean database, recently developed in the offshore oil and gas
industry, can be of great value.
Considerable research is required to develop reliable means of computing the extreme
response of offshore wind energy converters. Structural reliability methods recently
developed in the field of advanced offshore engineering should be extended to the
extreme loading of offshore wind energy converters, especially during production.
Loading by (near-)breaking waves in shallow waters and breaking sea ice loads should
be investigated to achieve safe but not overly conservative structures.
The use of the wind turbine control system of pitch-regulated designs for lowering of the
aerodynamic and hydrodynamic fatigue loads should be studied. Also the investigation
of active or passive mass dampers could be worthwhile for reducing the lateral support
structure response [11.1, 11.2].
References
247
REFERENCES
[1.1]
[1.2]
[1.3]
[1.4]
[1.5]
[1.6]
[1.7]
[1.8]
[1.9]
[1.10]
[1.11]
[1.12]
[1.13]
[1.14]
[1.15]
[1.16]
BTM Consult ApS, 2000, International Wind Energy Development - World
Market Update 1999, Forecast 2000 - 2004. (March 2000).
Matthies, H. et al., 1995, Study on Offshore Wind Energy in the EC (JOUR
0072). Verlag Natürliche Energien, Brekendorf.
Lemming, J., 1997, The Danish 4000 MW Offshore Programme. Proc. EWEC
1997, 37 - 42.
Gaudiosi, G. (ed.), 1994, Offshore Wind Energy in Mediterranean and other
European Seas - Technology and Implementation (OWEMES 1994), Proc.,
ENEA, Rome.
European Seas - Technology and Implementation (OWEMES 1997), Proc., 2nd
ed., ENEA, Rome.
European Seas - Technology and Implementation (OWEMES 2000), Proc.,
ENEA, Rome.
Machielse, L.A.H. (ed.), 1998, Identificatie van de Probleemgebieden bij de
Ontwikkeling van Grote Offshore Windturbines, ECN-CX–98-077, (in Dutch).
Watson, G., 1999, Research Requirement Workshop 12 April 1999 - Final
Report, Offshore Wind Energy Network, Rutherford Appleton Laboratory.
SEAS (ed.), 1997, Action Plan for Offshore Wind Farms in Danish Waters.
Haslev, Denmark.
NOVEM, 1997, Haalbaarheidsstudie demonstratieproject nearshore wind park.
Executive Summary, NOVEM, Utrecht, (in Dutch).
Kühn, M.; Bierbooms, W.A.A.M. et al., 1999, Towards a Mature Offshore Wind
Energy Technology - Guidelines from the Opti-OWECS Project. Wind Energy,
2, 25 - 58.
Hendriks, H. B.; van Bussel, G. J. W. et al., 2000, DOWEC Concept Study Evaluation of Turbine Concepts for Large Scale Offshore Application, Proc.
OWEMES 2000, 211 - 219.
Quarton, D. C.; Wastling, M. A., 1993, Wind and Wave Loading of Offshore
Wind Turbines, Proc. EUWEC 1993, Bedford, H.S. Stephens & Associates,
575 - 578. (Final report in: [1.2])
Kühn, M., 1994, Simulation of Offshore Wind Turbines under Stochastic
Loading, Proc. EWEC ‘94, Tsipouridis, J.L. (ed.), Thessalonki, Markedonia,
Greece. 3, 845 - 851.
Germanischer Lloyd, 1995, Germanischer Lloyd Rules for Regulations IV Non-Marine Technology, Part 2 Offshore Wind Energy, ‘Regulation for the
Certification of Offshore Wind Energy Conversion System’.
Lensink-van der Veen, A. et al., 1997, Near Shore Wind Energy - Technische
Realisatie - Eindrapport. Stork Engineering Consultancy B.V. Rep.Nr. 61860SEC-002, (in Dutch).
248
[1.17]
[1.18]
[1.19]
[1.20]
[1.21]
[1.22]
[1.23]
[1.24]
[1.25]
[1.26]
[1.27]
[1.28]
[1.29]
[1.30]
[1.31]
[1.32]
[1.33]
References
Fuglsang, P.; Thomsen, K., 1998, Cost Optimization of Wind Turbines for
Large-Scale Offshore Wind Farms. Risø National Laboratory, Roskilde,
Denmark.
Tong, K.C.; Quarton, D.C.; Standing, R, 1993, FLOAT - A Floating Offshore
Wind Turbine System. Proc. 15th BWEA Wind Energy Conference 1993,
Mechanical Engineering Publications, London.
Ferguson, M.C. (ed.); Kühn, M. et al., 1998, Opti-OWECS Final Report Vol. 4:
A Typical Design Solution for an Offshore Wind Energy Conversion System,
Institute for Wind Energy, Delft University of Technology.
Siegfriedsen, S.; Böhmeke, G.; 1998, Multibrid Technology - A Significant Step
to Multi-Megawatt Wind Turbines, Wind Energy, 1, 2, 89 - 100.
Elsamprojekt (ed.), 1997, Vindmøllerfundamenter i havet. Slutrapport.
Elsamprojekt A/S, Rep.Nr. EP97/215, (in Danish, English summary in: Juhl, H.
et al., 1997, Cost Efficient Foundation Structures for Large Scale Offshore
Wind Farms. Proc. OWEMES’97, Sardinia, Italy.)
Kühn, M., 1996, Dynamics of Offshore Wind Energy Converters on a Monopile
Foundation - Design Study based on the First Dutch Offshore Wind Farm Lely,
Institute for Wind Energy, Delft University of Technology (IW 96.093R).
van der Wekken, A. J. P., 1995, Frequency Measurements NedWind 40 Wind
Turbines of Wind Fram Lely, ECN, Rep.Nr. ECN-CX-95-031.
Simpson, P. et al., 1994, Offshore Wind Energy Assessment Phase CII:
Reassessmant of Costs of Offshore HAWT’s and their Effect on Cost of
Energy, ETSU W/35/00200/REP, UK.
Diedrichs, V. et al., 1995, Offshore-Windenergiesysteme; Fördervorhaben des
BMFT Nr. 0329645 A und B; Vulkan Eng., Bremen, (in German).
RES (Renewable Energy Systems Ltd.), 1993, Feasibility Study for a Prototype
Offshore Wind Turbine, Volumes 1-3, Report of UK Government DTI ETSU
Contract no. W/35/00251/REP/A-C.
Oscar, D. S.; Paez, T. L., 1988, Analysis of Wind Turbines on Offshore
Support Structures Exited by Random Wind and Random Waves.
Albuquerque, New Mexico, USA, Sandia National Laboratories. Sandia Report
SAND87-1689.UC-60. (Shorted version in: 7th ASME Wind Energy Symposium,
1988, Swift, A.H.P., Thresher, R.W. (eds.) ASME, New York, 165 - 172.)
Sinclair, F. M.; Clayton, B. R., 1989, Excitation and Damping Forces on
Offshore Wind Turbines, In: Wind Engineering 13, 6, 276 - 292. (Shorted
version in: The Effect of Wave Loading on Offshore Wind Turbines. Proc.
EWEC 1989, London, Peregrinus Ltd., 1058 - 1063.)
Halfpenny, A., 1998, Dynamic Analysis of both On and Offshore Wind Turbines
in the Frequency Domain. PhD Thesis. Department of Mechanical Engineering.
University College London.
Henderson, A.R., 2000, Development of Analysis Tools for Large Floating
Offshore Wind Farms, PhD Thesis, University College London.
Lindenburg,C.; van Engelen, T.G., 2000, Tools for Integrated Time and
Frequency Domain Analysis of Offshore Wind Turbines. Proc. OWEMES 2000,
525 - 539.
personal communication with Stig Øye, Lyngby University, Denmark, 2000.
personal communication with Risø National Lab., Roskilde, Denmark, 2000.
References
[1.34]
[1.35]
[1.36]
[1.37]
[1.38]
[1.39]
[2.1]
[2.2]
[2.3]
[2.4]
[2.5]
[2.6]
[2.7]
[2.8]
[2.9]
[2.10]
[2.11]
249
Frandsen, S. (ed.); Chacón, A. et al., 1996, Measurements and Modelling of
Offshore Wind Farms, Final Report JOU-CT93-0350. Risø National Lab.,
Roskilde.
Ronsten, G.; Thor, S.-E. et al., 2000, Evaluation of Loads, Power Quality, Grid
Interaction, Meteorological Conditions and Power Performance of the first
Swedish Offshore Wind Farm at Bockstigen, Proc. OWEMES 2000. 253 - 267.
Quarton, D.C., et al., (forthcoming), Design Methods for Offshore Wind
Turbines at Exposed Sites. Final Report JOR3-CT98-0284.
Kühn, M.; 2000, Design of the First Generation of Commercial Offshore Wind
Energy Converters - Taking the Enron Wind 2.0 (MW) Offshore and the
Utgrunden Project as an Example. Proc. DEWEK 2000, 234 - 237.
Bongers, P.M.M., 1994, Modelling and Identification of Flexible Wind Turbines
and Factorizational Approach to Robust Control, PhD Thesis, Delft University
of Technology.
de Ridder, H.A.J., 1994, “Design and Construct” of Complex Civil Engineering
Systems; a New Approach to Organisation and Contracts, PhD Thesis, Delft
University Press, Delft.
Kühn, M. (ed.), 1997, JOULE Proposal PL 971358: Concerted Action on
Offshore Wind Energy in Europe. Institute for Wind Energy, Delft University of
Technology, IW-97.131R.
Christensen, P.; Jørgensen, K.K. et al., 2000, Grid Connection and Remote
Control for the Horn’s Reef 150 MW Offshore Wind Farm in Denmark, Proc.
OWEMES 2000, 285 - 295.
Jamieson, P.; Quarton, D.C., 1999, Technology Development for Offshore,
Proc. EWEC 1999, 289 - 293.
Kühn, M.; Bierbooms, W.A.A.M. et al., 1999, Towards a Mature Offshore Wind
Energy Technology - Guidelines from the Opti-OWECS Project. Wind Energy,
2, 25 - 58.
Tong, K.C.; Quarton, D.C.; Standing, R, 1993, FLOAT - A Floating Offshore
Wind Turbine System. Proc. 15th BWEA Wind Energy Conference 1993,
Mechanical Engineering Publications, London.
Halfpenny, A. et.al., 1995, A Technical Feasibility Study and Economic
Assessment of Offshore Floating Wind Farm, Proc. BWEA 1995.
Coelingh, J.P.; A.J.M. van Wijk et. al, 1995, Large-Scale Wind Energy
Application - Transporting Wind Energy over Long Distances using an HVDC
Transmission Line in Combination with Hydro Energy or Biomass Energy.
Department of Science, Technology and Society, Utrecht University.
van Bussel, G.J.W., 1999, Development of an Expert System for the
Determination of Availability and O&M Costs for Offshore Wind Farms. Proc.
EWEC 1999, 402 - 405.
250
[2.12]
[2.13]
[2.14]
[2.15]
[2.16]
[2.17]
[2.18]
[2.19]
[2.20]
[2.21]
[2.22]
[2.23]
[2.24]
[2.25]
[2.26]
[2.27]
[2.28]
[2.29]
[2.30]
[2.31]
[2.32]
References
Quarton, D.C., et al., (forthcoming), Design Methods for Offshore Wind
Turbines at Exposed Sites. Final Report JOR3-CT98-0284.
SEAS (ed.), 2000, Design Regulations for Offshore Wind Farms. 1st ed. (in
Danish).
Heronemus, W.E., 1972, Power from the Offshore Winds. Proc. 8th Annual
Conf. Marine Techn. Soc., 435 – 466.
Beurskens, H.J.M., 2000, Offshore Wind Turbine Systems in Europe. ECN.
ECN-RX–00-001.
Hardell, R., 1990, Review Development in Offshore Wind Energy System
Technology to Mid 1990 and Look Forward. In: Energy and Environment into
the 1990's Proc. 1st Renewable Energy Congress, Reding. Sayigh, A.A.M.
(ed.); Pergamon Press, 1733-1742.
Bakker, C., et al., 1997, Winds of Change - A Feasibility Study of 10,000 MW
Offshore Wind Energy in the Nord Sea. E-Connection, Delft.
Kühn, M.; 2000, Design of the First Generation of Commercial Offshore Wind
OWEMES 2000, 211 - 219.
Hutting, H.K.; Cleijne, J.W. , 1999, The Price of Large Scale Offshore Wind
Energy in a Free Electricity Market, Proc. EWEC 1999. 399 - 402.
Beyer, H.G. et al., 1999, Forecast of Regional Power Output of Wind Turbines,
Proc. EWEC 1999, 1070 - 1073.
Cockerill, T.T.; Harrison, R.; Kühn, M. et al., 1998, Opti-OWECS Final Report
Vol. 3 - Comparison of Offshore Wind Energy at European Sites. Institute for
Wind Energy, Delft University of Technology.
Zervos, A., 2000, European Targets - Time to be More Ambitious. In: Wind
Directions. XIX, 6.
EWEA., 1999, A Plan for Action in Europe: Wind Energy - The Facts.
European Commission.
Grastrup, H.; Gaarde, J.K., 2000, Environmental Impact Assessment of the
First Four Offshore Wind Farms in Denmark, Proc. OWEMES 2000,. 417 - 427.
Annon., 2000, Milieu-Effectrapport, Locatiekeuze Demonstratieproject Near
Shore Windpark, Ministeries van EZ en VROM (in Dutch).
Erdmann, N.; Nanninga, F., 2000, Erfahrungen bei der Plaung und
Projektierung eines Offshore Windparks in der Nordsee. Proc. DEWEK 2000.
191 - 194 (in German).
Morthorst, P.E., 2000, Policy Instruments for Regulating the Development of
Wind Power in a Liberated Electricity Market. Proc. EWEC 1999. 7 - 12.
annom., 2001, New Energy, No. 2, p. 25.
annom., 2001, Geplante Offshore-Projekte in Deutschland, Neue Energie, No.
2 (In German).
European Commission, 1997, Energy for the Future: Renewable Sources of
Energy. White Paper for a Community Strategy and Action Plan. COM (97),
599 final.
References
[2.33]
[3.1]
[3.2]
[3.3]
[3.4]
[3.5]
[3.6]
[3.7]
[3.8]
[3.9]
[3.10]
[3.11]
[3.12]
[3.13]
[3.14]
[3.15]
[3.16]
[3.17]
251
Lemming, J.; Trong, M.D., 1997, Danish Investigations and Plan for Action for
Offshore Wind Power, Proc. OWEMES 1997, 189 - 199.
Kühn, M. (ed.) et al., 1998, Opti-OWECS Final Report Vol.1 - Integrated
Design Methodology for OWECS, Institute for Wind Energy, Delft University of
Technology.
Quarton, D.C., 1998, The Evolution of Wind Turbine Design Analysis - A
Twenty Year Progress Review, Wind Energy, 1, 5 - 24.
personal communication with Enron Wind GmbH, 2000.
Langen, van, M., 1997, Cost Effective Support Structure for an Offshore Wind
Energy Converter System (OWECS), Master Thesis, Workgroup Offshore
Technology, Delft University of Technology.
OWEMES 2000, 285 - 295.
Hutting, H.K.; Cleijne, J.W. The Price of Large Scale Offshore Wind Energy in
a Free Electricity Market, Proc. EWEC 1999. 399 - 402.
de Ridder, H.A.J., 1994, “Design and Construct” of Complex Civil Engineering
Systems; a New Approach to Organisation and Contracts, PhD Thesis, Delft
University Press, Delft.
Madsen, P.S., 1996, Tunø Knob Offshore Wind Farm, Proc. EUWEC ‘96.
Zervos, A. et al. (eds.); H.S. Stephens & Associates, Bedford, 4 - 8.
Olsen, F.; Rasmussen, K., 1994, Experience from the Construction and
Operation of Vindeby Offshore Wind Farm, Proc. OWEMES 1994.
Kouwenhoven, H.J. et.al., 1994, Wind Farm Lely - The First Offshore Wind
Farm in the Netherlands, Proc. EWEC 1994, Tsipouridis, J.L. (ed.), 3, 280 283.
Månsson, K.J., 1991, Pilot Plant for Offshore Windpower. Proc. EWEC 1991.
Van Hülle, Smulders, Dragt ( eds.), Elsevier Science Publishers B.V., 446-448.
Vølund, P.; Hansen, J., 2000, Middelgrunden 40 MW Offshore Wind Farm near
Copenhagen, Denmark, Installed Year 2000, Proc. OWEMES 2000, 369 - 378.
Cockerill, T.T. (ed) et.al., 1998, Opti-OWECS Final Report Vol. 2: Methods
Assisting the Design of Offshore Wind Energy Conversion Systems - Part A:
Economic Comparison of Design Options, Institute for Wind Energy, Delft
University of Technology.
Schöntag, C.; van Bussel, G.J.W., 1998, Opti-OWECS Final Report Vol. 2:
Methods Assisting the Design of Offshore Wind Energy Conversion Systems Part B: Optimisation of Operation and Maintenance. Institute for Wind Energy,
Delft University of Technology.
OWEMES 2000, 211 - 219.
Kühn, M., 2000, Design of the First Generation of Commercial Offshore Wind
252
[3.18]
[3.19]
[3.20]
[3.21]
[3.22]
[3.23]
[4.1]
[4.2]
[4.3]
[4.4]
[4.5]
[4.6]
[4.7]
[4.8]
[4.9]
[4.10]
[4.11]
[4.12]
[4.13]
[4.14]
References
Grievink, J.; Smit, K. et al, 1993, Managing Reliability and Maintenance in the
Process Industry. Proc. 2nd Int. Conf. Foundations of Computer-Aided Process
Operations. Rippin, D.W.T.; Halle, J.C.; Davis, J.F. (editors). Crested Butte,
Colorado, July 18-23, 1993, 133-158.
Siegfriedsen, S.; Böhmeke, G.; 1998. Multibrid Technology - A Significant Step
to Multi-Megawatt Wind Turbines. Wind Energy, 1, 89 - 100.
Deijl, van T.J.; Vries, de E., 1996, MultiWind Turbine; Energiecentrale voor de
21ste eeuw. (in Dutch, with English summary).
Ferguson, M.C. (ed.); Kühn, M.; et. al., 1998, Opti-OWECS Final Report Vol.
4: A Design Solution for an Offshore Wind Energy Conversion Systems.
Department of Energy, 1990, Offshore Installations: Guidance on Design,
Construction and Certification, London, HMSO, 4th ed.
Stull, R.B., 1988, An Introduction to Boundary Layer Meteorology, Dordrecht,
Kluwer.
van der Hoven, I., 1957, Power Spectrum of Horizontal Wind Speed in a
Frequency Range from 0.0007 to 90 Hz, Jour Meteorology, 14, 160 - 164.
Coelingh, J.P.; Folkerts, L. et al., 2000, Study on the Offshore Wind Climate
using a SODAR, Proc. OWEMES 2000, 61 - 70.
Barthelmie, R.J., 2000, Measurements and Modelling of Coastal Meteorology,
Proc. OWEMES 2000, 45 - 59.
Højstrup, J., 1999, Vertical Extrapolation of Offshore Wind Profiles. Proc.
EWEC 1999. James & James, London, 1220 - 1223.
Lange, B.; Højstrup, J., 1999, Influence of Waves on the Offshore Wind
Resources, Proc. EWEC 1999, James & James, London, 1216 - 1219.
Smedman, A.-S.; Bergström, H.; Bergström, U., 1995, Spectra, Variances and
Length Scales in a Stable Marine Environment Dominated by a Low Level Jet.
Boundary-Layer Meteorol., 66, 105 - 126.
Charnock, H., 1955, Wind Stress on a Water Surface, Quarterly Journal of the
Royal Meteorological Society, 81, 639.
van Wijk, A.J.M.; Beljaars, A.C.M. et al., 1990, Evaluation of Stability
Corrections in Wind Speed Profiles over the North Sea, J. of Wind Eng. Ind.
Aerodynamics. 33, 551 - 566.
Gasch, R. et al., 1996, Windkraftanlagen. 3rd ed. Teubner, Stuttgart (in
German).
DS 472 Danish Standard, 1992, DS 472 - Loads and Safety of Wind Turbine
Construction, 1st edition May 1992, English translated by NEL.
Germanischer Lloyd, 1995, Rules for Regulations IV - Non-Marine Technology,
Part 2 Offshore Wind Energy, Regulation for the Certification of Offshore Wind
Energy Conversion System’, Germanischer Lloyd, Hamburg.
Germanischer Lloyd, 1998, Rules and Regulations, IV - Non-marine
Technology, Part 1 - Wind Energy, Regulation for the Certification of the Wind
Energy Conversion Systems. Chapter 1 - 10. 1993 Edition incl. Supplement 1
March 1994 and Supplement 2 March 1998.
References
[4.15]
[4.16]
[4.17]
[4.18]
[4.19]
[4.20]
[4.21]
[4.22]
[4.23]
[4.24]
[4.25]
[4.26]
[4.27]
[4.28]
[4.29]
[4.30]
[4.31]
[4.32]
[4.33]
[4.34]
253
IEC, 1998, IEC 61400-1 ed. 2: Wind Turbine Generator Systems - Part 1:
Safety Requirements. Draft for Voting.
Shinozuka, M.; Jan, C.-M., 1972, Digital Simulation of Random Processes and
its Applications. Jour. Sound Vibration, 25, 1, 111 - 128.
Veers, P.S., 1998, Three Dimensional Wind Speed Simulation, Technical
Report SAND88-0152, Sandia National Lab.
Winkelaar, D., 1992, SWIFT - Program for Three-Dimensional Wind
Simulation; Part 1: Model Description and Program Verification, Techn. Report
ECN-R-92-013, ECN.
Dragt, J.B., 1990, Atmospheric Turbulence Characteristics in the Rotating
Frame of Reference of a WECS Rotor. Proc. EWEC 1990, H.S. Stephens and
Associates, 274 - 278.
Bierbooms, W.A.A.M; Dragt, J.B., 1994, The Choice of Atmospheric
Turbulence Models for Stochastic Rotor Loads Calculations. Proc. EWEC 1994
Tsipouridis, J.L. (ed.) Thessalonki, Macedonia, Greece, 1, 648 - 653.
personal communication with M. Rees, aerodyn Energiesysteme GmbH, 6th
February 1996.
Cleijne, J.W.; Coelingh J.P. et al., 1991, Description of the North Sea Wind
Climate for Offshore Wind Energy Application, TNO Environment and Energy
Research, Apeldoorn.
Pryor, S.C.; Barthelmie, R.J., 2000, Flow Characteristics in the Coastal Zone.
Proc. OWEMES 2000. 29 - 43.
Larson, G.C.; Jørgensen, H.E., 1999, Design Off-Shore Wind Climate. Proc.
EWEC 1999. James & James, London. 1038 - 1041.
Frandsen, S. (ed.); Chacón, A. et al., 1996, Measurements and Modelling of
Offshore Wind Farms, Final Report JOU-CT93-0350. Risø National Lab.,
Roskilde.
Adams, B.M. (ed); et al. 1996. Dynamic Loads in Wind Farms II, Final Report
Joule Project JOU2-CT92-0094, Garrad Hassan and Partners, Bristol.
Frandsen, S.; Thøgersen, M.L., 2000, Integrated Fatigue Loading for Wind
Turbines in Wind Farms, Proc. OWEMES 2000, 545 - 555.
CEN/TC 250/SC 1, 1992, EUROCODE 1, Basis of Design and Actions on
Structures, Part 2.7, Wind Actions. Draft of the Project Team PT 5.
DIBT (Deutsches Institut für Bautechnik), 1993, Richtlinie für Windkraftanlagen
- Einwirkungen auf Turm und Standsicherheitnachweise für Turm und
Gründung. Berlin, June 1993 (in German).
Bierbooms, W.A.A.M.; Cheng, P.-W. et al., 1999, Modelling of Extreme Gusts
for Design Calculations - NewGust, Proc. EWEC 1999, James & James,
London, 1001 - 1004.
Harland, L.A.; Vugts, J.H. et al., 1996, Extreme Response of Non-Linear
Dynamic Systems using Constrained Simulations, Proc. 15th Conference on
Offshore Mechanics and Arctic Engineering (OMAE), I-A, 193 - 200.
Holthuijzen, L.H., 1992, Wind en Golven. Proc. Symposium Wind Technologie
Buitengaats, 18 November 1992, Apeldoorn, KIVI and NEWIN.
Chakrabarti, S. K., 1987, Hydrodynamics of Offshore Structures, Southampton,
Computational Mechanics Publications, Springer.
Goda, Y., 1990, Random Waves and Spectra. In: Herbich, J.B. (Ed.) Handbook
of Coastal and Ocean Engineering. Vol. 1 - Wave Phenomena and Coastal
Structures, Gulf, Houston.
254
[4.35]
[4.36]
[4.37]
[4.38]
[4.39]
[4.40]
[4.41]
[4.42]
[4.43]
[4.44]
[4.45]
[4.46]
[4.47]
[4.48]
[4.49]
[4.50]
[4.51]
[4.52]
[4.53]
[4.54]
[4.55]
References
Brebbia, C.A.; Walker, S., 1979, Dynamic Analysis of Offshore Structures,
London, Butterworth & Co.
Pierson, W.J.; Moskowitz, L., 1964, A Proposed Spectral Form of Fully
Developed Wind Seas based upon Similarity Theory of S.A. Kitaigorodskii. J.
Geophys. Res., 69.
Hasselman, K. et al. 1973, Measurements of Wind Wave Growth and Swell
Decay during the JONSWAP Project. Deut. Hydrog. Z., Suppl. A, 8, 12, 95ff.
Barltrop, N.D.P.; Adams, A.J., 1991, Dynamics of Fixed Marine Structures, 3rd
ed., Butterworh-Heinemann Ltd., Oxford.
Wheeler, J.D., 1970, Method of Calculating Forces Produced by Irregular
Waves, Jour. of Petroleum Techn., 22, 359-367.
Dean, R.G., 1965, Stream Function Representation of Non-Linear Ocean
Waves, J. Geophs. Res., 70, 4561 - 4572.
Galvin, C.J., 1968, Breaker Type Classification on Three Laboratory Beaches.
J. Geophs. Res., 73, 3651 - 3659.
New, A.L.; McIver, P.; Peregrine, D.H., 1985, Computations of Overturning
Waves, J. Fluid Mech., 150.
Cokelet, E.D.; 1977, Breaking Waves, Nature, 267, 769 - 774.
Basco, D.R.; Niedzweck, J.M., 1989. Breaking Wave Force Distribution and
Design Criteria for Slender Piles, Proc. OTC 6009, 425 - 431.
DS 449, 1983, Pile-Supported Offshore Steel Structures Part 2, Annexes A-F,
Supplementray Guide, 1st ed October 1983.
Borgman, L.E, 1980, Techniques for Computer Simulation of Ocean Waves.
In: Topics in Ocean Physics. P. M. Rizzoli (ed.). Amsterdam NHDC.
Bouyssy, V.; Naboishikov, S M.; Rackwitz, R., 1993, Comparison of Analytical
Counting Methods for Gaussian Processes, Structural Safety, 12, 1, 35-57.
SIPM, 1993, Practice - Fatigue Analysis of Steel Substructures for Fixed
Offshore Platforms. Shell Internationale Petroleum Maatschappij B.V. (SIPM),
The Hague. Report EP 93-2005.
Hofland, L.D., 1999, Fatigue of the Support Structure of an OWEC due to
Combined Loading of Wind and Waves, Master Thesis, Section Offshore
Haapanen, E.; Määttänen, M.; Koskinen, P., 1997, Offshore Wind Turbine
Foundations in Ice Infested Waters, Proc. OWEMES 1997, 297 - 304.
Elsamprojekt (ed.), 1997, Vindmøllerfundamenter i havet. Slutrapport.
Elsamprojekt A/S, Rep.Nr. EP97/215, (in Danish). (English summary in: Juhl,
H. et al., 1997, Cost Efficient Foundation Structures for Large Scale Offshore
Wind Farms. Proc. OWEMES’97, Sardinia, Italy.)
Gravesen, H.; Sørensen, C., et al., 1999, Research Memo for Design Bases
for Ice Forces at the Middelgrunden, Carl Bro A/S, Denmark (in Danish).
Bundesamt für Seeschiffahrt und Hydrographie, 1996, Ostsee-Handbuch - Teil
B: Naturverhältnisse, Bundesamt für Seeschiffahrt und HydrographieNr. 2003,
Hamburg 18. Nov. 1996 (in German).
Svensson, P.; Thor, S.-E., 1992. Development Study III - Appendix 4.1.1
External Conditions and Load Cases. Vattenvall AB and FFA Sweden.
References
[4.56]
[4.57]
[4.58]
[4.59]
[4.60]
[4.61]
[4.62]
[4.63]
[4.64]
[4.65]
[5.1]
[5.2]
[5.3]
[5.4]
[5.5]
[5.6]
[5.7]
[5.8]
255
Hallam, M.G., Heaf, N.J., Wootton, L.R., 1977, Dynamics of Marine Structures:
Methods of Calculating the Dynamic Response of Fixed Structures Subject to
Wave and Current Action, London, Ciria Underwater Engineering Group.
Neumann, G.; Pierson, W.J., 1963, Principles of Physical Oceanography.
Prentise Hall.
Watson, G. (ed.), 1999, Offshore Wind Resources Assessment and Metocean
Data, Final Report of the OWEN Workshop 8th November 1999, Rutherford
Appleton Lab, www.owen.org.uk
Peters, D.J., Shaw, C.J., et al., 1993, Modelling the North Sea through the
North European Storm Study, Proc. OTC 7130, Houston, Texas.
Coelingh, J.P. (ed.), et al, 2000, Wind and Wave Data Compiled for the
DOWEC Concepts Study. DOWEC-T56-JC-00-R01_F-P.doc, Delft Univ. of
Techn., Section Wind Energy, IW-00162R
Cheng, P.-W.; van Bussel, G.J.W., 2000. Extreme Loading of Offshore Wind
Turbines Using Constrained Simulations, Proc. OWEMES 2000, 151 - 162.
Simmer, K.,1987, Grundbau - Teil 1: Bodenmechanik und erdstatische
Berechnungen, 18th ed., Teubner Stuttgart (in German).
Arbeitskreises 9 “Baugrunddynamik” der Deutschen Gesellschaft für Erd- und
Grundbau, 1992, Empfehlung E1 Bodendynamische Kennwerte,. Bautechnik,
69, 9 (in German).
NEN 6740, Geotechniek, TGB 1990 – Basiseisen en Belastingen
(Geotechnics, TGB 1990 – Basic Requirements and Loads), Nederlands
Normalisatie Instituut, ISBN 90-5254-074-8, Delft, December 1991.
(Wijzigingsblad NEN 6740/A1, Mei 1997), (in Dutch).
Andersen, O.V., 2000, Design Regulations for Offshore Wind Farms.
Specification for Concrete and Steel Structures, Proc. OWEMES 2000, 187 194.
of Technology.
Zaaijer, M.B., 1999, DUWECS Reference Guide. Institute for Wind Energy,
Delft University of Technology.
Bierbooms, W.A.A.M., 1990, A Dynamic Model of a Flexible Rotor - Part 1:
Description of the Mathematical Model, Institute for Wind Energy, Delft
University of Technology (IW 89.034 R).
Kühn, M., 1999, DUWECS Reference Guide - Errata V 3.4. Institute for Wind
Energy, Delft University of Technology.
Øye, S., 1996, Flex4 - Simulation of Wind Turbine Dynamics. Proc. 28th IEA
Meeting of Experts “State of the Art of Aeroelastic Codes for Wind Turbine
Calculations”. Techn. Univ. of Denmark, Lyngby, 71-76.
Cheng, P.-W., 1997, Optimization of Rotor Blades of an Offshore Wind
Turbine, Master Thesis, Institute for Wind Energy, IW97.111R, Delft University
of Technology.
Rodenburg, R.V., 1994, FOCUS - A Design for Structural Optimization of Rotor
Blades. Proc. EWEC 1994, Tsipouridis, J.L. (ed.), Thessalonki, Markedonia,
Greece. 3, 173 - 177.
Quarton, D.C., 1998, The Evolution of Wind Turbine Design Analysis - A
Twenty Year Progress Review, Wind Energy, 1, 5 - 24.
256
[5.9]
[5.10]
[5.11]
[5.12]
[5.13]
[5.14]
[5.15]
[5.16]
[5.17]
[5.18]
[5.19]
[5.20]
[5.21]
[5.22]
[5.23]
[5.24]
[5.25]
[5.26]
[5.27]
[5.28]
[5.29]
[5.30]
[5.31]
References
Snel, H.; van Kuik, G.A.M., 1998, Wind Turbine Rotor Aerodynamics: Problems
and Attempts to Solutions, Proc. 4th Eccomas Computational Fluid Dynamics
Conf., Athens, Volume with additional papers, 11 p.
Freris, L.L. (ed.), 1990, Wind Energy Conversion Systems, Prentice Hall Int.
Spera, D.A. (ed.), 1994, Wind Turbine Technology: Fundamental Concepts of
Wind Turbine Engineering, American Society of Mech. Eng.
Eggleston, D.M.; Stoddard, F.S., 1987, Wind Turbine Engineering Design, New
York, Van Nostrand Reinhold Company.
Molly, J.-P., 1990, Windenergie - Theorie, Anwendung, Messung, 2nd ed., C.F.
Müller, Karlsruhe (in German).
Gasch, R. et al., 1996, Windkraftanlagen. 3rd ed. Teubner, Stuttgart (in
German).
Sørensen, J.N., 1998, Status and Perspectives on CFD in Wind Energy, Proc.
4th Eccomas Computational Fluid Dynamics Conf., Athens, III, 442 - 449.
Froude, R.E., 1889, On the Part Played in Propulsion by Difference of Fluid
Pressure. Transactions Inst. Naval Architects., 30, 390 pp.
Lancaster, F.W., 1915, A Contribution to the Theory of Propulsion and the
Screw Propeller, Trans. Inst. Naval Arch., XXX, 330.
Betz, B., 1920, Das Maximum der theoretisch möglichen Ausnützung des
Windes durch Windmotoren. Zeitschrift für das gesamte Turbinenwesen, Sept.
20th, 1920, 207 - 209 (in German).
Glauert, H., 1963, Windmills and Fans, In: Durand, W.F. (ed.) Aerodynamic
Theory. 4, Div. L, Chap. XI, Dover Publications Inc, New York.
Abbott, I.H.; von Doenhoff, A.E., 1959, Theory of Wing Sections. Dover Publ.,
New York.
Prandtl, L.; Betz, A., 1927, Vier Abhandlungen zur Hydrodynamik und
Aerodynamik, Göttingen, 68 pp (in German).
Schmitz, G, 1955/56, Theorie und Entwurf von Windrädern optimaler Leistung,
Wissenschafliche Zeitungs der Universität Rostock, 5. Jahrgang (in German).
Wilson, R.E.; Lissaman, P.B.S.; et al., 1976, Aerodynamic Performance of
Wind Turbines, ERDA/NSF/04014-76/1, Washington D.C., US Depart. of
Energy.
Molenaar, D.-P., 1999, State-of-the-Art of Wind Turbine Design Codes. N-525,
Fac. Design, Eng. and Prod., Delft Univ. of Technology.
Snel, H; Houwink, R. et al., 1994, Sectional Prediction of Lift Coefficients on
Rotating Wind Turbine Blades in Stall, ECN-C-93-052.
Schepers, G.; Snel, H., 1995, Dynamic Inflow: Yawed Conditions and Partial
Span Pitch Control, ECN-C-95-056.
Snel, H., 1998, Review of the Present Status of Rotor Aerodynamics. Wind
Ener., 1, Pilot Issue, 46 - 69.
Björck, A., et al., 1995, Dynamic Stall and Three-Dimensional Effects, Final
Report JOU2-CT93-0345, FFA TN 1995-31.
Petersen, J.T. et al., 1998, Prediction of Dynamic Loads and Induced
Vibrations in Stall, Risø-R-1145(EN).
Wefelmeier, K., 1993, Dynamic Models of Wind Turbine Systems for Design
Purposes - Electrical Conversion Schemes, Final Report JOUR 0110,
Technical Univ. Braunschweig.
Heier, S., 1999, Grid Integration of Wind Energy Conversion Systems, John
Wiley & Sons.
References
[5.32]
[5.33]
[5.34]
[5.35]
[5.36]
[5.37]
[5.38]
[5.39]
[5.40]
[5.41]
[5.42]
[5.43]
[5.44]
[5.45]
[5.46]
[5.47]
[5.48]
[5.49]
[5.50]
[5.51]
[5.52]
[5.53]
257
Feijoo, A.E.; Cidras, J., 1999, A Dynamic Model for Doubly-Fed Asynchronous
Wind Turbines. Proc. EWEC 1999, James & James, London. 897 - 900.
Hansen, A.D., Bindner, H., et al, 1999, Improving Transition between Power
Optimization and Power Limitation of Variable Speed / Variable Pitch Wind
Turbines. Proc. EWEC 1999, James & James, London. 889 - 982.
Krüger, T., 1999, Regelungsverfahren von WindKraftanlagen zur Reduktion
von mechanischen Belastungen. PhD Thesis, GH Kassel (in German).
Bossanyi, E.A.; Jamieson, P., 1999, Blade Pitch System Modelling for Wind
Turbines. Proc. EWEC 1999, James & James, London. 893 - 896.
Kühn, M., 1993, Dynamics of an Offshore Wind Turbine - Digital Simulation.
Annon., 1992, ANSYS 5.3 - User’s Manual. Swanson Analysis Systems Inc.
Schaumann, P.; Seidel, M., 2000, Eigenschwingungsverhalten von
Windenergieanlagen - Berechnungen und Messungen. Proc. DEWEK 2000,
93 - 96 (in German).
Franz, G., (ed.), 1978, Beton-Kalender 1978, Teil II. Vol. 67. Wilhelm Ernst &
Sohn, Düsseldorf (in German).
Randolph, M.F., 1981, The Response of Flexible Piles to Lateral Loading.
Géothechnique. 31, 2, 247 - 259.
API RP2A-LRFD, 1993, API Recommended Practices for Planning, Designing
and Constructing Fixed Offshore Platforms - Load and Resistance Factor
Design. First Edition, July 1, 1993.
Verruijt, A., 1994, Offshore Soil Mechanics. Lecture Notes x2-ct2. Fac. Civil
Eng. Delft Univ. of Technology.
Savidis, S.A. (ed.), 1998, Empfehlungen des AK 1.4 ‘Baugrundynamik’ der Dt.
Gesellschaft für Geotechnik - Teil E3 Federsteifigkeiten und Dämpfungen auf
dem Baugrund. Bautechnik, 75, 10, 792 - 805 (in German).
Arbeitskreises 9 “Baugrunddynamik” der Deutschen Gesellschaft für Erd- und
Grundbau, 1992, Empfehlung E1 Bodendynamische Kennwerte,. Bautechnik,
69, 9 (in German).
Richart, F.E.; Hall, J.R.; Woods, R.D., 1971, Vibration of Soils and
Foundations, Prentice-Hall, Englewood Cliffs, New York.
Roesset, J.M.; Whitman, R,V.; at al, 1973, Modal Analysis of Structures with
Foundation Interaction. J. Struc. Div., ASCE, 99, ST3, March, 399 - 416.
Wolf, J.P., 1994, Foundation Vibration Analysis Using Simple Physical Models.
Prentice Hall.
DIN 4133. Steel Stacks. November 1991.
DIBT (Deutsches Institut für Bautechnik), 1995, Richtlinie für Windkraftanlagen - Einwirkungen auf Turm und Standsicherheitnachweise für Turm und
Gründung. Berlin. 2th ed. (in German).
Chakrabarti, S. K., 1987, Hydrodynamics of Offshore Structures, Southampton,
Computational Mechanics Publications, Springer.
Morison, J.R.; O’Brien, M.P. at al., 1950, The Force Exerted by Surface Waves
on piles. Petroleum Transactions, AIME. 189, 149-157.
258
[5.54]
[5.55]
[5.56]
[5.57]
[5.58]
[5.59]
[5.60]
[5.61]
[5.62]
[6.1]
[6.2]
[6.3]
[6.4]
[6.5]
[6.6]
[6.7]
[6.8]
[6.9]
[6.10]
[6.11]
References
MacCamy, R.C.; Fuchs, R.A, 1954, Wave Forces on Piles: A Diffraction
Theory. Tech. Memo 69., US Army Corps of Engineers, Beach Erosion Board.
Hofland, L.D., 1999, Description and Evaluation of Hydrodynamic Model in
DUWECS. Section Offshore Technology. Delft University of Technology.
Kärnä, T., 1993, Steady State Vibrations of Offshore Structures. Proc. 1st
International Conf. Development Russian Actic Offshore, 20 - 24 Sept.,
St.Petersburg.
DS 410, 1998, Codes for Construction Loads, Danish Standard, 4th edition.
Ralston, T.D., 1979, Plastic Limit Analysis of Sheet Ice Loads on Conical
Structures, International Union of Theoretical and Applied Mechanics
Symposium on Physics and Mechanics of Ice, Copenhagen, Denmark.
Gravesen, H.; Sørensen, C., et al., 1999, Research Memo for Design Bases
for Ice Forces at the Middelgrunden, Carl Bro A/S, Denmark (in Danish).
Xu, J., 1993, Aeroelastik einer Windturbine mit drei gelenkig befestigten
Flügeln, VDI Fortschritssberichte, Reihe 11 Nr. 185 (in German).
Garrad, A.D, 1990, Forces and Dynamics of Horizontal Axis Wind Turbines. In:
Freris, L.L. (Ed.), Wind Energy Conversion Systems. Prentice Hall Int., 119 144.
575 - 578. (Final report in: [1.2])
Greece. 3, 845 - 851.34
Oscar, D. S.; Paez, T. L., 1988, Analysis of Wind Turbines on Offshore
Support Structures Exited by Random Wind and Random Waves.
Albuquerque, New Mexico, USA, Sandia National Laboratories. Sandia Report
SAND87-1689.UC-60. (Shorted version in: 7th ASME Wind Energy Symposium,
1988, Swift, A.H.P., Thresher, R.W. (eds.) ASME, New York, 165 - 172.)
Kortlever, W., 1992, The Use of the Discrete Element Method to Model an
Offshore Wind Turbine. Master Thesis. Delft University of Technology.
Garrad Hassan & Partners., 1996, Bladed for Windows. Theory Manual. GH
Report 282/BR/009, October 1996.
Lindenburg, C.; van Engelen, T.G., 2000, Tools for Integrated Time and
Frequency Domain Analysis of Offshore Wind Turbines. Proc. OWEMES 2000,
525 - 539.
personal communication with Stig Øye, Lyngby University, Denmark, 2000.
personal communication with Risø National Lab., Roskilde, Denmark, 2000.
Bierbooms, W.A.A.M., 1990, A Dynamic Model of a Flexible Rotor - Part 1:
Description of the Mathematical Model, Institute for Wind Energy, Delft
University of Technology (IW 89.034 R).
Holten, van, T., 1987, Hamiltionian Mechanics as a Possible Alternative for the
Deriving Aeroelastic Equations. Proc. 13th European Rotorcraft Forum. Arles,
France.
References
[6.12]
[6.13]
[6.14]
[6.15]
[6.16]
[6.17]
[6.18]
[6.19]
[6.20]
[6.21]
[6.22]
[6.23]
[6.24]
[6.25]
[6.26]
[6.27]
[6.28]
[6.29]
[6.30]
259
Kühn, M., 1993, Dynamics of an Offshore Wind Turbine - Digital Simulation.
Eggleston, D.M.; Stoddard, F.S., 1987, Wind Turbine Engineering Design, New
York, Van Nostrand Reinhold Company.
Petersen J.T., 1990, Kinematically Nonlinear Finite Element Model of a
Horizontal Axis Wind Turbine, Ph.D. Thesis, Risø Nat. Laboratory.
Patel, M.; Wei, J. et al., 1993, Further Development of a Finite Element Code
for Use by the Wind Turbine Industry, ETSU W/43/00147/REP, Harwell.
Link, M.; Nowak, B., 1986, Zur dynamischen Analyse des Systems
Stromabnehmer und Fahrleitung, dargestellt am Beispiel des Intercity
Experimental. In: Dynamik gekoppelter Systeme, VDI-Bericht 603, VDI Verlag
Düsseldorf, 245 - 261 (in German).
Kaiser, K., 2000, Luftkraftverursachte Steifigkeits- und Dämpfungsmatrizen von
Windturbinen und ihr Einfluß auf das Stabilitätsverhalten. VDI
Fortschrittsberichte Reihe 11, Nr. 294 (in German).
Peinelt, I., 1995, Untersuchung des Einflusses der luftkraftverursachten
Dämpfungs- und Steifigkeitsmatrizen auf das dynamische Verhalten einer
Windturbine. Master Thesis. Institute for Aerospace Engineering, TU Berlin (in
German).
Nedergaard, H., Tychsen, J., et al. Ringing and Double Frequency Response
of a Tripod, Transactions of the ASME, Jour Offshore Mech. & Arctic Eng, 119,
79 - 85.
Vugts, J.H.; Kinra, R.K., 1976, Probabilistic Fatigue Analysis of Fixed Offshore
Structures, Proc. OTC 2608.
Quarton, D.C., Wastling, et al., 1992, The Calculation of Wind Turbine Loads.
A Frequency or Time Domain Problem? Proc. Wind Energy Conversion 1992.
Clayton, B.R. (ed.), Mech. Eng. Publ., London, 307 - 314.
Gasch, R.; Knothe, K., 1989, Strukturdynamik Bd. 2 - Kontinua und ihre
Diskretisierung, Springer (in German).
Vugts, J.H. et al., 1979, Modal Superposition versus Direct Solution
Techniques in the Dynamic Analysis of Offshore Structures, Proc. BOSS ‘79.
BHRA Fluid Eng., Cranfield, Bedford, UK.
Guyan, R.J., 1965, Reduction of Stiffness and Mass Matrices,. AIAA-J 3, 2.
Hurty, W.C., 1965, Dynamic Analysis of Structural Systems Using Component
Modes, AIAA-J 3, 678 - 685.
of Technology.
Wallrapp, O., 1994, Standardization of Flexible Body Modeling in Multibody
System Codes - Part I: Definition of Standard Input Data. Mech. Struct. &
Mach. 22, 3, 283 - 304.
Kurz, W., 1994, Finite Element Methods in a Simulation Code for Offshore
Wind Turbines, Institute for Wind Energy, Delft Univ. of Technology.
260
[6.31]
[6.32]
[7.1]
[7.2]
[7.3]
[7.4]
[7.5]
[7.6]
[7.7]
[7.8]
[7.9]
[7.10]
[7.11]
[7.12]
[7.13]
[7.14]
[7.15]
[7.16]
[7.17]
References
Bierbooms, W.A.A.M; Dragt, J.B., 1994, The Choice of Atmospheric
Turbulence Models for Stochastic Rotor Loads Calculations. Proc. EWEC 1994
Tsipouridis, J.L. (ed.) Thessalonki, Macedonia, Greece, 1, 648 - 653.
van Delft, D. R. V.; van Leeuwen, J. L., 1994, Full Scale Fatigue Testing of
Wind Turbine Rotor Blades. Proc. EWEC 1994 Tsipouridis, J.L. (ed.)
Thessalonki, Markedonia, Greece. 1, 727 - 737.
Kaiser, K., 2000, Luftkraftverursachte Steifigkeits- und Dämpfungsmatrizen von
Windturbinen und ihr Einfluß auf das Stabilitätsverhalten. VDI
Fortschrittsberichte Reihe 11, Nr. 294 (in German).
of Technology.
Hovde, G.O.; Moan, T., 1997, Fatigue and Overload Reliability of Mooring
Systems, Proc. Int. Offshore and Polar Engineering Conf., 4, Int Soc of
Offshore and Polar Engineers (ISOPE), Golden, CO, USA.,145-152.
Lie, H.; Fylling, I.J., 1994, Evaluation of Methods for Fatigue Analysis of
Offshore Mooring Lines, Proc. Offshore South East Asia, 10th Conference and
Exhibition, Singapore, 345 - 352.
Jiao, G.; Moan, T., 1990, Probabilistic Analysis of Fatigue due to Gaussian
Load Processes, Probabilistic Engineering Mechanics, 5, 2, 76 - 83.
Wirsching, P.H.; Light, M.C., 1980, Fatigue under Wide Band Random
Stresses, Journal of the Structural Division, ASCE, 106, ST7, July, 1593-1607.
Jiao, G., 1989, Stochastic Analysis of Combined Dynamic Responses and their
Effects. Computational Mechanics Proc. Asian Pac. Conf. Comput. Mech. A.
Balkema, Roterdam, 671 - 676.
Dirlik, T., 1985, Application of Computers in Fatigue Analysis, PhD Thesis,
Univ. of Warwick.
Bishop, N.W.M.; Zhihua, H., 1991, The Fatigue Analsyis of Wind Turbine
Blades Using Frequency Domain Techniques. Proc. EWEC, 1991. Van Hülle,
Smulders, Dragt ( eds.); Elsevier Science Publishers, 246 - 250.
Lambert, R.G., 1988, Fatigue Damage Accumulation Prediction for Combined
Sine and Random Stresses, Jour. of Environmental Sciences, 31, 3, p. 53-63.
Sakai, S.; Okamura, H., 1995, On the Distribution of Rainflow Ranges for
Gaussian Random Processes with Bimodal PSD, JSME Int Journal Series ISolid Mechanics Strength of Materials, 38, 4, 440-445.
personal communication with R. Fraser, AMEC - Advanced Eng. Services,
Aberdeen, UK, 1999.
Rice, S. O., 1952, Mathematical Analysis of Random Noise, Selected Papers
on Noise and Stochastic Processes, N. Wax ed., Dover, New York.
Vanmarcke, E.H., 1972, Properties of Spectral Moments with Applications to
Random Vibration, Jour. Engineering Mechanics Division, ASCE, 98, EM2,
April, p 425 - 445.
Madsen, P.H. (ed), 1990, Recommended Practices for Wind Turbine Testing
and Evaluation: 3. Fatigue Loads. 2nd ed. Submitted to the Exec. Committee
References
[7.18]
[7.19]
[7.20]
[7.21]
[7.22]
[7.23]
[7.24]
[7.25]
[7.26]
[7.27]
[7.28]
[8.1]
[8.2]
[8.3]
[8.4]
[8.5]
[9.1]
261
of the IEA Programme for Research and Development of Wind Energy
Conversion Systems.
Miner, M.A., 1945, Cumulative Damage in Fatigue, Journal Applied Mechanics,
12.
Paris, P.C.; Erdogan, F, 1963, A Critical Analysis of Crack Propagation Laws,
Int. J. Fatigue, 9, 2.
van Grol, H.J., Bulders, B.H., et al., 1994, Reference Procedure to Establish
Fatigue Stresses for Large Wind Turbines: A State of the Art Report Vol.1 Erik
H., : Main report and Annexes, ECN-C--94-013, Netherlands Energy Research
Foundation.
van Delft, D.R.V.; van Leeuwen, J.L., 1994, Full Scale Fatigue Testing of Wind
Turbine Rotor Blade, Proc. European Wind Energy Conference 1994, 1, p.
727-737.
Vugts, J.H., Kinra, R.K., 1976, Probabilistic Fatigue Analysis of Fixed Offshore
Structures. Proc. OTC 2608.
Chaudhury, G.K.; Dover, W.D.,1985, Fatigue Analysis of Offshore Platforms
Subject to Sea Wave Loading. Int. J. Fatigue. 7, 1, 13 - 19.
Hancock, J.W.; Gall, D.S., 1985, Fatigue under Narrow and Broad Band
Stationary Loading, Final Report of the Cohesive Programme of Research and
Development into Fatigue of Offshore Structures, Marine Technology
Directorate Ltd. UK.
Chow; Li, 1991, Analytical Solution for Fast Fatigue Assessment under WideBand Random Loading, Int. Jour. Fatigue,. 13, 5, 395 - 404.
Kim; Kim, 1994, Simple Method for Evaluation of Fatigue Damage of
Structures in Wide-Band Random Vibrations, Proc. Inst. Mech. Eng. Part C Jour. of Mech. Eng. Science. 208, 1, 65 - 68.
Lui; Pengmin, 1997, Analytical Solution of Equivalent Stress for Structure
Fatigue Life Prediction under Broad band Random Loading, Mechanics of
Structures and Machines, 25, 2, 139 - 149.
Bouyssy, V.; Naboishikov, S.M.; Rackwitz, R., 1993, Comparison of Analytical
Counting Methods for Gaussian Processes, Structural Safety, 12, 1, p 35-57.
Peters, D.J., Shaw, C.J., et al., 1993, Modelling the North Sea through the
North European Storm Study, Proc. OTC 7130, Houston, Texas.
van Bussel, G.J.W., 1999, Development of an Expert System for the
Determination of Availability and O&M Costs for Offshore Wind Farms. Proc.
EWEC 1999, 402 - 405.
SIPM, 1993, Practice - Fatigue Analysis of Steel Substructures for Fixed
Offshore Platforms. Shell Internationale Petroleum Maatschappij B.V. (SIPM),
The Hague. Report EP 93-2005.
IEC, 1998, IEC 61400-1 ed. 2: Wind Turbine Generator Systems - Part 1:
Safety Requirements. Draft for Voting.
Germanischer Lloyd, 1995, Rules for Regulations IV - Non-Marine Technology,
Part 2 Offshore Wind Energy, Regulation for the Certification of Offshore Wind
Energy Conversion System’, Germanischer Lloyd, Hamburg.
262
[9.2]
[9.3]
[9.4]
[9.5]
[9.6]
[9.7]
[9.8]
[9.9]
[9.10]
[9.11]
[9.12]
[9.13]
[9.14]
[9.15]
[9.16]
[9.17]
[9.18]
[9.19]
[10.1]
[10.2]
[10.3]
References
Kühn, M., 1996, Interpretation and Background Information on Modal Analysis
and Dynamic Simulations of the Offshore Wind Farm Lely. Institute for Wind
Energy IW96.109R, Delft University of Technology.
Annon., 1994, Windmolen fundering Medemblik. Ballast-Nedam Engineering.
Projectnr. 193126. Rev. 01 (in Dutch).
Kühn, M., 1995, Modal Analysis of the NedWind 40 Tower – Comparison of
Measurements and Calculations for the Sites Oostburg and Lely (IJsselmeer),
Institute for Wind Energy IW96.092R, Delft University of Technology.
Kaverner Turbin AB, 1997, Development Study III - Final Report. Kristinehamn.
575 - 578. (Final report in [1.2])
Greece. 3, 845 - 851.
Magnus, K., 1986, Schwingungen. 4th ed. Teubner, Stuttgart.
Schaumann, P.; Seidel, M., 2000, Eigenschwingungsverhalten von
Windenergieanlagen - Berechnungen und Messungen. Proc. DEWEK 2000,
93 - 96 (in German).
Savidis, S.A. (ed.), 1998, Empfehlungen des AK 1.4 ‘Baugrundynamik’ der Dt.
Gesellschaft für Geotechnik - Teil E3 Federsteifigkeiten und Dämpfungen auf
dem Baugrund. Bautechnik, 75, 10, 792 - 805 (in German).
van der Wekken, A.J.P., 1995, Acceleration Measurements NedWind 40
Turbines of Wind Farm Lely. ECN-CX–95-031.
personal communication with M.B. Zaaijer, TU Delft, November 2000.
Kühn, M. (ed.), 1998, Validatie van OWEC ontwerpgereedschappen en
aanbevelingen voor ontwerpmethodes. Projectvoorstel TWIN II (In Dutch).
Kühn, M., 1998, Simplified Analysis of a Monopile Support Structure for the TW
1.5s at a Baltic Site. Institute for Wind Energy. IW-98149R.
Vugts, J.H.; Dob, S.L. et al., 1998, The Extreme Dynamic Response of Botton
Supported Structures Using an Equivalent Quasi-Static Design Wave
Procedure. Applied Ocean Research, 20, 37 - 53.
Ferguson, M, et al., 1996, Concept Development Offshore Wind Energy
Conversion Sub-Systems. Technical Report Opti-OWECS Project Phase II.
Kühn, M., 1997, Dynamics of Offshore Wind Energy Converters - Contribution
to the JOULE III Project Opti-OWECS. Wind Engineering, 21, 3, 161 - 180.
Kühn, M., 1997, Soft or Stiff - A Fundamental Question for Designers of
Offshore Wind Energy Converters. Proc. EWEC 1997. 575 - 578.
Neumann, G.; Pierson, W.J., 1963, Principles of Physical Oceanography,
Prentise Hall.
References
[10.4]
[10.5]
[10.6]
[10.7]
[10.8]
[10.9]
[10.10]
[10.11]
[10.12]
[10.13]
[10.14]
[10.15]
[10.16]
[10.17]
[10.18]
[10.19]
[10.20]
[10.21]
[10.22]
263
Bundesamt für Seeschiffahrt und Hydrographie, 1996, Ostsee-Handbuch - Teil
B: Naturverhältnisse, Bundesamt für Seeschiffahrt und Hydrographie Nr. 2003,
Hamburg (in German).
Annon., 1991, Richtline zur Auslegung, Aufstellung und das Betreiben von
Windkraftanlagen, Fassung Januar 1991, Amtsblatt für Schleswig-Holstein,
Kiel, Nr. 10 (in German).
Adam, J.H., 1982, Longman Dictionary of Business English, Harlow Longman.
Tande, J.0.; Hunter, R., 1994, Recommended Practices for Wind Turbine
Testing: 2. Estimation of Cost of Energy from Wind Energy Systems, IEA, 2nd
ed.
EWEA., 1999, A Plan for Action in Europe: Wind Energy - The Facts.
European Commission.
Köhler, S., 1995, Cost Analysis and Optimisation of Offshore Wind Energy
Converter Systems, Institute for Wind Energy, Delft University of Technology
IW-95.088R.
Pauling, T., 1996, Economic Optimization of Offshore Wind Farms. Institute for
Wind Energy, Delft University of Technology IW-96.098R.
Kühn, M., et al., 1996, Cost Analysis and Optimization of Offshore Wind
Farms, Proc. EUWEC 1996, H.S. Stephens & Associates, 134 - 137.
Cockerill, T.T., et al., 1998, Opti-OWECS Final Report Vol. 2: Methods
Assisting the Design of OWECS. Part A: Development of a Cost Model for
Offshore Wind Energy, Institute for Wind Energy, Delft University of
Technology.
Greece. 3, 845 - 851.
SEAS, NNR, Rambøll, DGI and Risø, 1998, Wind Turbine Foundations in the
Sea. EFP-96. Development project supported by the Danish Energy Agency.
(in Danish). (English summary in: Juhl, H., et al., 1997, Cost Efficient
Foundation Structures for Large Scale Offshore Wind Farms. Proc. OWEMES
1997. 397 - 410.)
Annon., 1991, A Study of Offshore Based Wind Power. The Blekinge Project
Summary Report, Sweden.
OWEMES 2000, 285 - 295.
Freris, L.L. (ed.), 1990, Wind Energy Conversion Systems, Prentice Hall Int.
Harland, L.A.; Vugts, J.H. et al., 1996, Extreme Response of Non-Linear
Dynamic Systems using Constrained Simulations, Proc. 15th Conference on
Offshore Mechanics and Arctic Engineering (OMAE), I-A, 193 - 200.
Cheng, P.-W.; van Bussel, G.J.W., 2000, Extreme Loading of Offshore Wind
Turbines Using Constrained Simulations, Proc. OWEMES 2000, 151 - 162.
API RP2A-LRFD, 1993, API Recommended Practices for Planning, Designing
and Constructing Fixed Offshore Platforms - Load and Resistance Factor
Design. First Edition, July 1, 1993.
264
[11.1]
[11.2]
[A.1]
References
Kareem, A.; Tamura, Y, 1994, Damping Systems for Controlling Wind Induced
Motions of Structures. Proc. 12th ASCE Structures Congress, Atlanta, GA,
Baker, N.C.; Goodno, B.J. (eds.), I, 845 - 850.
Enevoldsen, I.; Mork, K.J., 1996, Effects of a Vibration Mass Damper in a Wind
Turbine Tower. Mechanics of Structures & Machines, 24, 2, 155 - 187.
Kühn, M., 1996, Terminology, Reference Systems and Conventions to be used
within the Opti-OWECS Project, Delft Univ. of Technology, Institute for Wind
Energy. Report IW 96.097R.
Offshore Wind Energy Terminology
265
APPENDIX A: OFFSHORE WIND ENERGY TERMINOLOGY
Offshore wind energy is a fairly young and multidisciplinary field. Unfortunately, no
uniform terminology exists and misunderstandings can occur quite easily. Therefore,
within the Opti-OWECS project one has agreed upon a particular terminology,
conventions and reference systems in order to make the internal and external
communication more effective. Here a shortened version of reference [A.1] is presented.
Preface
In principle, the common practice concerning notation and convention within the
considered disciplines i.e. wind engineering, offshore technology and engineering
economics should be used. However, harmonization is required in the description of the
entire system and its components, the interfaces between sub-systems and the
structural design. The two former aspects are treated in this appendix.
1 Offshore wind energy conversion system (OWECS)
offshore wind energy conversion system (OWECS)
Entire system, comprising (usually) several wind energy converter units, for
conversion of wind energy into electric power including the wind turbines, the
support structures, the grid connection to the power delivery point and operation
and maintenance aspects. Note that the environment i.e. air, water and soil as well
as the utility grid is not considered as a part of the OWECS.
offshore wind farm
Synonym for OWECS.
2 Subsystems of the OWECS
Subsystems of an OWECS (Figure A-1) comprise either physical part of the system, e.g.
wind turbine, support structure, or important aspects such as operation and maintenance
aspects.
OWECS (Offshore Wind Energy Conversion System, offshore wind farm)
/) offshore wind energy converter (OWEC)
*
/) wind turbine
*
.) support structure
/) grid connection
.) operation and maintenance aspects
Figure A.1: Subsystems of the (bottom-mounted) OWECS
266
Appendix A
offshore wind energy converter (OWEC)
Single unit of the OWECS comprising wind turbine and support structure. 1
wind turbine (WT)
Component of an offshore wind energy converter that transforms wind energy into
electric power on generator voltage or AC-rectifier voltage, comprising rotor, nacelle
with entire interior, control and safety system and electrical turbine system.
support structure (bottom-mounted)
Structure that supports the wind turbine and transfers the loading into the soil.
Hence, the support structure comprises both the tower and the foundation.
grid connection and wind farm layout
This comprises two main parts that are considered for convenience as one
subsystem. Firstly, the electrical system that takes the power provided at the turbine
connection points and collects it at the wind farm collection point(s) and
successively transmits it to the onshore connection point with the public grid.
Secondly the physical arrangement of the OWEC units.
operation and maintenance aspects
Auxiliary facilities, equipment and strategy required for operation, maintenance,
control and administration of an OWECS
3 Boundaries of OWECS components
wind turbine and support structure
The fixed end of the yaw mechanism of the nacelle is defined as the boundary
between the (horizontal axis) wind turbine and the support structure. All geometric
and dynamic conditions are expressed with respect to the reference frame of the
support structure.
wind turbine and grid connection
The turbine switch gear or circuit breaker is defined as boundary between electrical
system of the turbine and the grid connection. The voltage at the connection point
corresponds to the generator or the inverter (if any). Although a transformer might
be installed at the wind energy converter unit, it is regarded as part of the grid
connection.
grid connection and utility grid
The power of the OWECS is provided as three-phase AC at the voltage level of the
utility grid to which the wind farm is connected. In absence of other explicit
conventions the connection point is situated at the first dry location onshore
regardless the actual grid infrastructure on land.
1
No plural of the abbreviation ‘OWEC’ should be used; instead one may use ‘OWEC units’, full
spelling or the singular form ‘OWEC’, if possible.
Offshore Wind Energy Terminology
The main components of the OWECS sub-systems are defined by Figure A.2.
OWECS
/) wind turbine
* /) rotor
* /) (mechanical) drive train
* /) nacelle enclosure
* /) control and safety system
* .) electrical turbine system
/) support structure
* /) tower
* .) foundation
/) grid connection and wind farm layout
* /) power collection
* .) power transmission
* * /) offshore
* * .) onshore
* .) wind farm layout
.) operation and maintenance aspects
/) maintenance facilities and equipment
/) control, safety and administration facilities
.) operation and maintenance strategy
Figure A.2: OWECS sub-components
267
268
Appendix B
APPENDIX B: DATA OF THE REFERENCE SITES AND DESIGNS
Table B.1: Main data of the reference sites in Chapter 9
Site
Wind
conditions
MSL
Hydrodynamic
conditions
Vave = 7.7 m/s
k=2
α = 0.14
hhub = 41.5 m
5m
Hmax = 1.77 m
Tass = 6.2 s
U = 0.5 m/s
surge = 0.2 m
Vave = 8.5 m/s
k=2
α = 0.11
hhub = 65 m
Vref = 42 m/s
10 m
Hs50 = 3.44 m
Tz50 = 7.6 s
Hmax = 6.4 m
Tass = 9.4 s
surge = 1.2 m
tice = 0.4 m
8300 km
medium dense
to dense sand,
gravel,
boulders, some
clay
NL-5
Vave = 10.1
m/s
k = 2.2
α = 0.11
hhub = 59 m
25 m
Hmax = 15.4 m
Tass = 12.5 s
U = 1.2 m/s
surge = 2.75 m
tide = 0.75 m
50 2000
km
weak to firm
clay
Blekinge
Vave = 8.3 m/s
k=2
α = 0.11
hhub = 51.5 m
15 m
Hmax = 9.7 m
Tass = 11 s
U = 0.4 m/s
surge = 0.8 m
tice = 0.6 m
7500 km
weak to firm
clay
Lely A3
Lely A2, A2mod
Utgrunden
(Turbine No. 4)
10 m
Fetch
0.8 32 km
Soil type
sand and clay
sand and clay,
slip at top layer
medium dense
11.5 Hs50 = 6.9 m
sand
2000
Tz50 = 7.7 s
km
Hmax = 12.8 m
Tass = 9.5 s
U = 0.8 m/s
surge = 3 m
tide = 1 m
tice = thickness of extreme 50 years sea-ice
α = wind shear exponent
k = Weibull shape factor
Tass = wave period associated with Hs50
Hmax = extreme 50 years wave heightTz50 = 50 year zero crossing period
U = 50 year design current
Hs50 = 50 year significant wave height
surge = storm surge
Vave = annual average wind speed at hhub
tide = extreme tidal amplitude
Vref = 50 year wind speed at hhub (10 min
mean)
NL-1
Vave = 9 m/s
k = 2.2
α = 0.11
hhub = 58.6 m
15 20 m
Data of the Reference Sites and Designs
269
Table B.2: Lumped load cases for six directional sectors at the NL-1 site
Sector NNE and SSW:
( 15o to 45o and 195o to 225o)
Sector N and S:
( -15o to 15o and 165o to 195o)
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
sum
Hs [m]
Tz [s]
0.57
0.86
0.74
0.82
0.99
1.12
1.28
1.41
1.56
1.78
1.95
2.09
2.36
2.66
2.81
3.10
3.79
2.75
3.92
4.83
4.00
3.92
4.19
4.25
4.40
4.51
4.61
4.76
4.89
5.00
5.17
5.27
5.42
5.68
6.26
5.25
V1h [m/s]
<5.0
<5.0
6.0
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18.5
19.5
20.5
21.5
Prob. [0/000]
200
91
316
139
132
109
111
72
49
42
31
27
16
12
10
4
3
1
Hs [m]
Tz [s]
V1h [m/s]
0.44
0.70
0.67
0.79
0.88
0.99
1.16
1.31
1.48
1.66
1.85
2.07
2.27
2.43
2.57
2.73
2.95
3.10
3.16
3.28
3.85
3.54
4.40
3.74
3.93
4.02
4.13
4.22
4.40
4.54
4.63
4.75
4.86
4.98
5.10
5.19
5.29
5.43
5.54
5.55
5.44
5.91
<5.0
<5.0
6.1
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18.5
19.5
20.5
21.5
22.5
>25.0
>25.0
1365
Sector NNE and SSW:
( 45o to 75o and 225o to 255o)
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
sum
Hs [m]
Tz [s]
0.49
0.82
0.70
0.81
0.91
1.02
1.22
1.43
1.59
1.79
1.96
2.20
2.38
2.61
2.80
2.97
3.17
3.41
3.53
3.60
4.37
3.58
4.35
3.74
3.80
3.89
3.99
4.28
4.48
4.60
4.73
4.83
4.97
5.12
5.21
5.34
5.46
5.58
5.69
5.80
5.67
6.33
V1h [m/s]
<5.
<5.
6.1
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18.5
19.5
20.5
21.5
22.5
>25.
>25.
Prob. [0/000]
175
43
293
169
207
198
226
193
161
144
129
123
83
71
51
34
31
17
8
8
9
2373
Prob. [0/000]
139
46
248
165
176
168
164
121
95
83
76
68
48
39
25
19
13
9
5
4
6
1717
Sector ENE and WSW:
( 75o to 105o and 255o to 285o)
Hs [m]
Tz [s]
0.49
0.80
0.72
0.86
0.99
1.10
1.28
1.46
1.64
1.86
2.08
2.31
2.53
2.73
2.98
3.19
3.41
3.73
3.86
4.10
5.19
3.55
4.37
3.78
3.86
4.09
4.18
4.36
4.54
4.65
4.80
4.94
5.11
5.24
5.35
5.52
5.65
5.76
6.01
6.03
6.12
6.90
V1h [m/s]
<5.
<5.
6.1
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18.5
19.5
20.5
21.5
22.5
>25.
>25.
Prob. [0/000]
106
47
222
132
146
161
152
137
129
95
87
74
53
37
30
27
19
15
9
11
9
1698
270
Appendix B
Table B.2 (cont.): Lumped load cases for six directional sectors at the NL-1 site
Sector ESE and WNW:
( 105o to 135o and 285o to 315o)
No.
Hs [m]
Tz [s]
V1h [m/s]
Prob. [0/000]
<5.
3.59
0.51
1
<5.
4.50
0.83
2
6.1
3.93
0.76
3
7.5
3.91
0.86
4
8.5
4.18
1.06
5
9.5
4.29
1.15
6
10.5
4.47
1.33
7
11.5
4.62
1.51
8
12.5
4.75
1.72
9
13.5
4.90
1.97
10
14.5
5.03
2.16
11
15.5
5.15
2.38
12
16.5
5.36
2.65
13
17.5
5.56
2.93
14
18.5
5.67
3.17
15
19.5
5.89
3.46
16
20.5
5.98
3.61
17
21.5
6.18
3.96
18
22.5
6.30
4.22
19
>25.
6.37
4.25
20
>25.
6.85
5.03
21
sum
V1h = hourly mean wind speed at 60 m height
77
80
214
122
135
125
135
90
71
55
47
38
31
27
20
15
12
9
7
4
6
1320
Sector SSE and NNW:
( 135o to 165o and 315o to 345o)
Hs [m]
Tz [s]
0.58
0.93
0.79
0.93
1.10
1.22
1.40
1.60
1.76
2.02
2.31
2.53
2.85
3.14
3.31
3.63
3.96
4.28
4.29
4.64
5.48
3.94
4.84
4.10
4.04
4.32
4.39
4.54
4.71
4.85
5.00
5.22
5.40
5.55
5.76
5.86
6.11
6.36
6.50
6.50
6.61
7.13
V1h [m/s]
<5.
<5.
6.1
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18.5
19.5
20.5
21.5
22.5
>25.
>25.
Prob. [0/000]
180
112
292
149
149
133
119
92
71
51
35
28
20
18
11
10
12
4
4
4
5
1499
271
Table B.3: Soil conditions
Site
Soil conditions and parameters
depth
[m]
0-2
2-4
4-7
7-9
9 - 17
17 - 19
19 - 24
24 - 25
Lely A2,
Lely A2-mod
0-2
2 - 3.5
3.5 - 10
10 - 11
11 - 12
12 - 14
14 - 18
18 - 19.5
19.5 - 23
Utgrunden †
γsub = 10 kN/m3, φ = 35o, K = 28 MPa/m
NL-5
Gdyn = 8.3 MPa, ν = 0.5
Blekinge
Gdyn = 8.3 MPa, ν = 0.5
NL-1
φ = 30o (P-y curves generated by Kværner Oil & Gas Ltd.)
Lely A3
‡
soil
clay
clay
sand
sand
sand
sand
clay
sand
γsat
[kN/m3]
15.5
16
20
20
20
20
19
20
Cu
[kN/m2]
10
15
100
-
φ
[o]
29
30
32
34
36
slib
clay
sand
sand
sand
sand
clay
sand
sand
13
15.5
19.5
20
20
20
19
20
20
1
8
100
-
29
31
30
31
32
36
†
average soil conditions as used for sensitivity analysis, softer soil assumed for limit state
analysis
‡
assumed in Chapter 9
γsat
γsub
Cu
φ
K
Gdyn
ν
= saturated unit weight
= submerged unit weight
= undrained shear strength of clay
= internal friction angle of sand
= modulus of initial subgrade reaction
= dynamic shear modulus
= Poisson ratio
272
Appendix B
Table B.4: Main support structure data
Design
Dimensions
Lely A3
monopile: Ø 3.7 m x 35 mm x 28 m, 20.5 m penetration
tower: Øbase 3.19 m x 12 mm, Øtop 1.9 m x 12 mm, 37.9 m length
Lely A2
monopile: Ø 3.2 m x 35 mm x 26 m, 13.5 m penetration
tower: see Lely A3
Lely A2-mod
monopile: Ø 3.2 m x 35 mm x 34 m, 20 m penetration,
double conical contraction between -4 m and 3 m (MSL) including
a cylindrical section Ø 2.5 m x 50 mm x 4 m
tower: see Lely A3, but stretched to 45.4 m length
Utgrunden
monopile: Ø 3 m x 45 mm x 33.7 m, 18.7 m penetration
transition piece: Ø 3.15 m x 9.8 m
tower: Øbase 3.15 m, Øtop 2.6 m, double conical section Ø 2.25 m,
54.55 m length
(Turbine No. 4)
Gravity
lattice tower
see Figure 9.4
triangle edge length
height
- 23 m (MSL)
50 m
- 7 m (MSL)
18 m
8.32 m
21 m (MSL)
59 m (MSL)
3m
total ballast 1000 t
Gravity
Monotower
see Figure 9.4
total ballast 3000 t
Opti-OWECS
Monopile
see Figure 9.4
monopile: Ø 3.5 m x 75 mm x 45 m, 25 m penetration
including transition between -10 m and -7 m (MSL) and upper
section Ø 2.5 m x 100 mm x 6 m
transition piece: Ø 3.8 m x 60 mm x 15 m with 5 m grouted joint
tower:
1st section Ø 2.8 m x 60 mm x 12 m
2nd section Ø 2.8 m x 32 mm x 12 m
3rd section Ø 2.8 m x 25 mm x 12 m
4th section Ø 2.8 m x 20 mm x 6 m
5th section 4.6 m long transition to Ø 3.5 m x 30 mm
46.6 m length
273
Table B.5: Main wind turbine data
Wind turbine
NedWind 40
Enron Wind
1.5o
Kværner Turbin
WTS 80
WTS 80 M
Rated power
500 kW
1.5 MW
3 MW
Rated wind
speed
15 m/s
12 m/s
14.15 m/s †
13.7 m/s ‡
Diameter
40.77 m
70.5 m
80 m
Rotor concept
two-bladed,
rigid hub,
up-wind,
active yaw
three-bladed,
rigid hub
up-wind,
active yaw
two-bladed,
rigid hub
up-wind,
active yaw
Tilt angle
5o
4o
3o
Rotor speed
32 rpm
11 - 20 rpm
19 rpm
22 rpm
Gear box
parallel
planetary/
parallel
planetary or
planetary/parallel
Control concept
constant speed,
active stall
variable-speed,
variable pitch
constant speed,
variable pitch
Generator
asynchronous
double-feed
asynchronous
asynchronous
Features
2 generators
adjustable
power factor
flexible, hydraulic
yaw system
83 t
132 t
141 t
34 t
Nacelle and
rotor weight
†
WTS 80, ‡ WTS 80 M (marine)
†
‡
†
‡
274
Appendix B
Table B.6: Input data for typical DUWECS model of the WTS 80 M turbine
Parameter
Value
Module : General (general data)
cut-in wind velocity
cut-out wind velocity
rated wind velocity
rated aerodynamic rotor power
generator angular velocity
rated tip-speed ratio
rotor angular velocity
communication time step
number of integration steps per interval
4th order Runge-Kutta integration scheme
Module : ROTOR2 (rotor & nacelle)
number of blade sections (blade elements)
airfoil used between 6 m # z < 26.4 m
airfoil used between 26.4 m # z < 32.2 m
airfoil used between 32.2 m # z < 40 m
dimensionless power loss factor
blade contour
chord [m]
radius [m]
3.524
5.000
3.672
5.500
4.038
6.752
4.400
10.252
4.312
13.752
3.400
17.252
2.735
20.752
2.300
24.252
1.950
27.752
1.600
31.252
1.300
34.752
1.000
38.252
0.830
39.500
0.700
40.000
5. m/s
25. m/s
13.7 m/s
3.785 kW
161.79 rad/s
6.73
2.2038 rad/s
0.075 s
3.
.true.
10
FFA W3 22.1%
transition from W3 22.1% to LS-1
LS-1 - 15 % by FFA
1
blade mass between hinge and tip
flapwise blade inertia with respect to hinge
lead-lagwise blade inertia with respect to hinge
blade centre of gravity with respect to hinge
dimensionless blade hinge offset in terms of radius R
rotational stiffness of blade flap spring
rotational stiffness of blade lead-lag spring
flap damping ratio as fraction of critical damping
lead-lag damping ratio as fraction of critical damping
inertia of rotor shaft
rotational stiffness of rotor shaft
rotational damping of rotor shaft
time constant of wake retardation
twist with respect to rotor plane [o]
13.40
13.40
13.40
10.00
6.96
4.86
3.48
2.55
1.66
0.65
-0.20
-0.86
-0.95
-1.00
6382. kg
4.48 @ 105 kgm2
4.48 @ 105 kgm2
12.488 m
0.15
1.039 @ 108 Nm/rad
2.513 @ 108 Nm/rad
1. %
2. %
2.5 @ 106 kgm2
7.5 @ 109 Nm/rad
5. @ 107 Nms/rad
4. @ R / V
275
Table B.6 (cont.): Input data for typical DUWECS model of the WTS 80 M turbine
Parameter
Module : ROTOR2 (rotor & nacelle)
air density
gravity acceleration
wind shear exponent for power law
yaw angle of undisturbed wind velocity
variation of yaw angle by sine law during simulation
nacelle tilt angle (hub up positive)
(inclined) distance between hub centre and tower axis
outer radius of tower for tower shadow model
dimensionless mass eccentricity of rotor
total nacelle mass (used only for steady tower top loads)
x-coordinate of nacelle centre of gravity
Module : TOWER32 (support structure)
number of used mode
number of modes used as dynamic degrees of freedom
structural damping ratio as fraction of critical damping
modified Pierson-Moskowitz spectrum
significant wave height
zero up-crossing period
number of wave components in spectral discreatisation
equidistant frequency spacing
minimum ratio of PSD at peak and at cut-in frequency
cut-off frequency for spectral discreatisation
refined spacing near fundamental eigenfrequency
odd number of refined frequencies around eigenfrequency
minimum of average frequency increment in spectral tail
ratio between max. and min. frequency increment at tail
linear wave theory with Wheeler stretching
wave diffraction according to MacCamy and Fuchs
gravity acceleration
water density
water depth
Module : TRANSS (gear box)
transmission ratio
equivalent drive train inertia at rotor side
Module : ASM1 (asynchronous generator)
slope of the torque curve in operating point
grid voltage
Module : ANCON (analog controller)
maximum pitch rate
Value
1.25 kg/m3
9.81 m/s2
0.11
30.o
.true.
3.o
4.908 m
1.6 m
0.001
141,000. kg
-2.45 m
10.
6.
0.5 %
.true.
3.81 m
5.92 s
100.
.false.
0.002
0.637 Hz = 4 rad/s
.true.
13.
0.025
3.1416
.true.
.true.
9.807 m/s2
1025. kg/m3
20. m
70.227
6.9 @ 106 kgm2
3618.5 Nms/rad
690. V
5. o/s
276
Appendix B
Table B.7: Pitch controller and actuator model of the WTS 80 M turbine
pitch angle Θ
Θo
Wind turbine
rotor speed Ω
Controller
-Ωo
PI-T1 controller with 1st order delay for pitch actuator characteristics
Kp ti s % Kp
H(s) '
ti ta s 2 % ti s
where:
H(s)
s
Kp
ti
ta
transfer function
complex frequency s
gain, linearly interpolated between
Kp = 1.6 s at V = 15 m/s and Kp = 2.6 s at V = 24 m/s
integration time constant, linearly interpolated between
ti = 0.65 s at V = 15 m/s and ti = 1.28 s at V = 24 m/s
actuator time constant, ta = 0.5 s
277
SAMENVATTING
Windenergie buitengaats zal een belangrijke duurzame energiebron worden voor
verschillende Noord-Europese landen. Uitgaande van kleine prototypeprojecten in de
jaren negentig, die de technische en principiële economische haalbaarheid aantoonden,
vindt nu een snelle ontwikkeling plaats. In het jaar 2000 zijn voor het eerst windenergieconverters in de megawatt klasse buitengaats geplaatst; vanaf 2002 zullen die
standaard gebruikt gaan worden voor grote windparken van 100 tot 160 MW. Op middel
tot lange termijn bestaan er plannen voor multi-megawatt turbines en windparken tot één
gigawatt vermogen.
Voor grootschalige exploitatie van windenergie buitengaats is de ontwikkeling van een
volwassen offshore windenergietechnologie van cruciaal belang. Basis voor een
dergelijke ontwikkeling zijn de ervaringen met de windenergietechnologie op het land,
offshore technologie en de bedrijfsvoering van grote elektriciteitscentrales.
Hoe kunnen de economie en de betrouwbaarheid van offshore windenergieconversiesystemen door analyse van de dynamica en een toegespitste
ontwerpmethodologie verbeterd worden?
Deze vraagstelling is het onderwerp van het proefschrift en wordt in vier hoofdbijdragen
beantwoord:
• Ontwikkeling en demonstratie van een geïntegreerde ontwerpmethologie voor
offshore windparken
• Synthese van berekeningsmethodes met een complexiteit toegepast op de
verschillende stappen in het ontwerpproces
• Onderzoek van de specifieke dynamica van offshore windenergie-converters en de
interactie met hun ontwerp
• Demonstratie van een optimalisatie door afstemming van het dynamisch gedrag.
Hieronder worden de belangrijkste aspecten van deze gebieden geschetst.
Na een algemene inleiding wordt de huidige stand van de techniek samengevat en
daaruit wordt de geïntegreerde ontwerpaanpak voor offshore windparken afgeleid. Deze
methodiek weegt alle technische componenten zoals wind turbine,
ondersteuningsconstructie en netaansluiting en installatie-, bedrijfsvoerings- en
onderhoudsaspecten bij de beslissingen tijdens het ontwerpproces. Interacties tussen
deelsystemen worden zo volledig als praktisch mogelijk meegenomen. De
ontwerpoplossing wordt door beoordeling van systeemparameters gevonden,
bijvoorbeeld energiekosten, dynamica van het gehele systeem, aanpassing aan de
specifieke locatie, kosten van installatie en beschikbaarheid van het park.
Voor de analyse van de dynamica van offshore windenergie-converters (die vast
verbonden zijn met de zeebodem) wordt een algemene methode opgesteld door
samenvoeging van de expertise op gebied van windenergie op land en de offshore
278
technologie. Daardoor is de analyse van de gelijktijdige responsie van het gehele
systeem op wind, golven, stroming en ijs mogelijk.
Deze niet-lineaire simulaties in het tijddomein zijn wel nauwkeurig maar te ingewikkeld
voor vermoeiingsanalyses in het vroege stadium van het ontwerpproces. Daarom zijn
er ook meer eenvoudige methodes ontwikkeld. De equivalente vermoeiingsbelastingen
uit tijddomeinsimulaties van de windresponsie van de offshore windenergie-converter
in rustig water worden opgeteld bij de resultaten van de lineaire spectrale analyse van
de golfresponsie van de ondersteuningsconstructie. Deze splitsing is toelaatbaar als de
aërodynamische demping van de paal in rekening gebracht wordt. Voor iedere taak op
zich zijn de standaard ontwerppakketten uit de windenergie- en offshore technologie
toereikend.
Voor certificatieberekeningen wordt nog steeds de voorkeur gegeven aan geïntegreerde,
niet-lineaire simulaties van de wind- en golfresponsie in het tijddomein. De spreiding van
de wind- en golfparameters resulteert in een, uit praktisch oogpunt, te hoog aantal
belastingsgevallen. Door middel van een verdere methode kan dat teruggebracht
worden tot een klein aantal karakteristieke belastingsgevallen.
Bij het onderzoek van de dynamica ligt de nadruk op de door wind en golfslag belaste
ondersteuningsconstructie, omdat de windturbine weinig door de hydrodynamische
excitatie beïnvloed wordt. De interactie tussen het ontwerp en het dynamisch gedrag
wordt aan de hand van zeven voorbeelden van offshore windenergie-converters,
waarvan er drie daadwerkelijk gebouwd zijn, onderzocht. Het nominale vermogen
varieert van 500 kW tot 3 MW en de omgevingscondities betreffen zowel een
beschermd binnenwater als ook locaties blootgesteld an het ruwe klimaat van de
Noordzee. De meest belangrijke aspecten zijn: beschrijving van de omgevingscondities,
trillingsmodes en frequenties, aërodynamische demping, onzekerheden van het
funderingsgedrag en responsie onder vermoeiing- en extreme condities.
Aanbevelingen worden gegeven voor de dynamische analyse en het ontwerpproces.
De ontwerpmethodologie, de analysemethodes en de inzichten in de dynamica worden
gedemonstreerd aan de hand van het stapsgewijs ontwerpen en optimaliseren van een
300 MW offshore windpark.
Toepassing tijdens het Europese onderzoeksproject Opti-OWECS en de realisatie van
het Utgrunden offshore windpark in de Oostzee toonden reeds de voordelen aan van
een geïntegreerde benadering van de dynamica en het ontwerpproces.
279
ZUSAMMENFASSUNG
Offshore-Windenergie wird eine wichtige erneuerbare Energiequelle für verschiedene
nordeuropäische Länder werden. Ausgehend von den kleinen Demonstrationsprojekten
in geschützten Gewässern, die in den Neunzigern die technologische und prinzipielle
wirtschaftliche Machbarkeit zeigten, findet zurzeit eine rasante Entwicklung statt. Im
Jahre 2000 wurden erstmals Offshore-Windenergieanlagen der Megawatt-Klasse
errichtet, die ab dem Jahr 2002 für große Windparks mit jeweils 100 - 160 MW
installierter Leistung vorgesehen sind. Mittel- bis langfristig bestehen Pläne für MultiMegawatt-Windenergieanlagen und Windparks mit einer Leistung von bis zu einem
Gigawatt.
Eine solche großtechnische Nutzung erfordert die Entwicklung einer ausgereiften
Offshore-Windenergietechnik, die auf den Erfahrungen der landgestützten
Windenergienutzung, der Offshore-Technik der Öl- und Gasindustrie und dem Betrieb
von Großkraftwerken aufbaut.
Wie lässt sich die Wirtschaftlichkeit und Zuverlässigkeit von OffshoreWindenergiesystemen durch Analyse des strukturdynamischen Verhaltens und eine
gezielte Entwurfsmethodik verbessern?
Diese Fragestellung ist Gegenstand der vorliegenden Dissertation und wird in vier
Hauptbeiträgen beantwortet:
· Entwicklung und Demonstration einer integrierten Entwurfsmethodik für OffshoreWindparks
· Synthese von Berechnungsverfahren für das strukturdynamische Verhalten von
Offshore-Windenergieanlagen, deren Komplexität an die verschiedenen Stadien des
Konstruktionsprozesses angepaßt sind
· Untersuchung des besonderen strukturdynamischen Verhaltens von OffshoreWindenergieanlagen und der Wechselbeziehung zu ihrer konstruktiven Ausführung
· Demonstration einer Entwurfsoptimierung durch Abstimmung des
strukturdynamischen Verhaltens
Im folgenden werden die wesentlichen Ergebnisse in diesen Bereichen skizziert.
Nach einer allgemeinen Einleitung wird der derzeitige Stand der Technik
zusammengefasst und daraus die integrierte Entwurfsmethodik für Offshore-Windparks
abgeleitet. Diese Methode berücksichtigt alle technischen Komponenten wie
Windturbine, Tragstruktur und Netzanbindung und sowie Aspekte der Errichtung, des
Betriebes und der Wartung bei Entscheidungen während des Konstruktionsprozesses.
Wechselwirkungen zwischen den Teilsystemen werden so vollständig mitgenommen,
wie dies praktisch durchführbar ist. Entwurfsentscheidungen werden durch Auswertung
von Systemkenngrößen gefunden, wie z.B. die durchschnittlichen
Energiegestehungskosten, das dynamisches Verhalten des Gesamtsystems, die
Anpassung an die Standortbedingungen, den Aufwand für die Errichtung und die
Verfügbarkeit des Windparks.
280
Zur Untersuchung von ortsfesten Offshore-Windenergieanlagen wird ein
Referenzverfahren aus den Erfahrungen der landgestützten Windenergietechnik und der
Offshore-Technik synthetisiert. Gleichzeitige Beanspruchungen aus Wind, Wellen,
Strömung und Eisgang können analysiert werden.
Das Verfahren der nicht-linearen Simulation im Zeitbereich ist genau, aber zu aufwändig
für Betriebsfestigkeitsuntersuchungen während des frühen Entwurfsstadiums. Daher
werden vereinfachte Methoden entwickelt. Die schädigungsäquivalenten
Beanspruchungen aus Simulationen der Windbeanspruchungen im Zeitbereich werden
kombiniert mit den Ergebnissen der linearen spektralen Analyse der
Wellenbeanspruchungen im Frequenzbereich. Diese getrennte Betrachtungen sind
zulässig, wenn die aerodynamische Dämpfung der Tragstruktur berücksichtigt wird. Für
beide Aufgaben liegen allgemein verfügbare Programme aus der Windenergie- bzw.
Offshore-Technik vor.
Die integrierte, nicht-lineare Simulation der gleichzeitigen Systemantwort auf Wind- und
Wellenanregung im Zeitbereich wird nach wie vor für Zertifizierungsberechnungen
vorgezogen. Die Streuung der Wind- und Wellenparameter führt jedoch in praktischer
Hinsicht in eine zu hohe Anzahl von Lastfällen. Deshalb wird eine weitere Methode zur
Ermittlung einer geringen Anzahl von charakteristischen Lastfallkombinationen
eingeführt.
Bei der Behandlung der Strukturdynamik wird der durch Wind und Wellen
beanspruchten Tragstruktur besondere Aufmerksamkeit gewidmet, da die Windturbine
durch die hydrodynamischen Anregungen wenig beeinflusst wird. Die
Wechselbeziehung zwischen dem Entwurf und dem strukturdynamischen Verhalten wird
anhand von sieben, davon drei realisierten, Beispielen von OffshoreWindenergieanlagen untersucht. Die Nennleistung variiert von 500 kW bis 3 MW und
die Umweltbedingungen reichen von einem geschützten Binnengewässer bis zu
exponierten Standorten in der Nordsee. Die wichtigsten offshore-spezifischen Aspekte
des dynamischen Verhaltens eines Entwurfes umfassen die Beschreibung der
Umweltbedingungen, die Schwingungsformen und Frequenzen, die aerodynamische
Dämpfung, die Unwägbarkeiten der Fundamenteigenschaften und das Systemverhalten
unter Ermüdungs- und Extrembedingungen. Es werden Empfehlungen für
strukturdynamische Untersuchungen und die konstruktive Gestaltung erarbeitet.
Die Entwurfsmethodik, die Analysemethoden sowie die von den Beispielanlagen
abgeleiteten Erkenntnisse werden durch die schrittweise Konstruktion und Optimierung
eines 300 MW Offshore-Windparks illustriert.
Anwendungen während des europäischen Forschungsprojektes Opti-OWECS und der
Verwirklichung des Utgrunden Offshore-Windparks verdeutlichten die Vorteile einer
integrierten Behandlung der Dynamik und des Konstruktionsprozesses.
281
CURRICULUM VITAE
Martin Kühn was born in Linne, now Bissendorf, Germany on 20th May 1963. After
passing high school in 1983 and working with mentally disabled people in a mechanical
workshop he was educated at Kromschröder AG, Osnabrück. He won the annual award
granted by the Chamber of Industry and Trade for the best-trainee-toolmaker in Lower
Saxony in1987.
In 1987 he started to study Mechanical Engineering at the Hannover University and
graduated with honours in Physical Engineering at the Technical University of Berlin in
1993. Main subjects included Applied Mechanics, Structural Dynamics and Wind
Energy.
A practical training of five months on design, assembly and erection of wind turbines
was received at Südwind Energiesysteme, Berlin. The ‘Evangelische Studienwerk
Villigst e.V.’ granted a scholarship from 1989 to 1993 which enabled the conduction of
the master thesis ‘Dynamics of an Offshore Wind Turbine - Digital Simulation’ at the
Institute for Wind Energy of Delft University of Technology in 1992/93.
From 1993 to 1999 he was employed as research assistant, lecturer, research scientist
and project manager at the Institute for Wind Energy and contributed to various national
and European research and consultancy projects on dynamics, design and economics
of offshore wind farms. In 1996/97 he coordinated the EU Joule III project Opti-OWECS
‘Structural and Economic Optimisation of Bottom-Mounted Offshore Wind Energy
Converters’ (JOR3-CT95-0087) and compiled the first proposal for the ‘Concerted Action
on Offshore Wind Energy in Europe’ which finally started in the year 2000. More than 35
scientific and technical publications in the field of offshore wind energy were written and
a poster award for the ‘best overall contribution of scientific contents and presentation'
was received at the EUWEC 1996 Conference.
Since 1999 he is employed at Enron Wind GmbH, Salzbergen, Germany as project
manager for offshore wind energy engineering and coordinated the design of the
company’s first offshore wind farm installed on the Utgrunden reef in 2000.
282

Microsoft Photo Editor - omslagfront.jpg

Transcription

Similar documents

click here - Korean-American Offshore Engineers Association

Son of Putin ally has £35m London house

• Power and Intensity • Doppler effect for (i) mechanical waves e.g.

WE INSPIRE PEOPLE TO GO PLACES

2015 MCYLFT Flyer - Young Life Matagorda County

7pm Where

Programme booklet 2015

Numerical CFD comparison of Lillgrund employing RANS

summary and key conclusions